Network Working Group P. Riikonen Internet-Draft draft-riikonen-silc-spec-07.txt 29 July 2003 Expires: 29 January 2004 Secure Internet Live Conferencing (SILC), Protocol Specification Status of this Memo This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC 2026. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet-Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html The distribution of this memo is unlimited. Abstract This memo describes a Secure Internet Live Conferencing (SILC) protocol which provides secure conferencing services over insecure network channel. SILC provides advanced and feature rich conferencing services with security as main design principal. Strong cryptographic methods are used to protect SILC packets inside the SILC network. Three other specifications relates very closely to this memo; SILC Packet Protocol [SILC2], SILC Key Exchange and Authentication Protocols [SILC3] and SILC Commands [SILC4]. Riikonen [Page 1] Internet Draft 29 July 2003 Table of Contents 1 Introduction .................................................. 3 1.1 Requirements Terminology .................................. 4 2 SILC Concepts ................................................. 4 2.1 SILC Network Topology ..................................... 4 2.2 Communication Inside a Cell ............................... 6 2.3 Communication in the Network .............................. 7 2.4 Channel Communication ..................................... 7 2.5 Router Connections ........................................ 8 3 SILC Specification ............................................ 9 3.1 Client .................................................... 9 3.1.1 Client ID ........................................... 9 3.2 Server .................................................... 10 3.2.1 Server's Local ID List .............................. 11 3.2.2 Server ID ........................................... 12 3.2.3 SILC Server Ports ................................... 12 3.3 Router .................................................... 13 3.3.1 Router's Local ID List .............................. 13 3.3.2 Router's Global ID List ............................. 14 3.3.3 Router's Server ID .................................. 14 3.4 Channels .................................................. 14 3.4.1 Channel ID .......................................... 16 3.5 Operators ................................................. 16 3.6 SILC Commands ............................................. 17 3.7 SILC Packets .............................................. 17 3.8 Packet Encryption ......................................... 17 3.8.1 Determination of the Source and the Destination ..... 18 3.8.2 Client To Client .................................... 19 3.8.3 Client To Channel ................................... 20 3.8.4 Server To Server .................................... 21 3.9 Key Exchange And Authentication ........................... 21 3.9.1 Authentication Payload .............................. 21 3.10 Algorithms ............................................... 23 3.10.1 Ciphers ............................................ 23 3.10.1.1 CBC Mode .................................. 24 3.10.1.2 CTR Mode .................................. 24 3.10.1.3 Randomized CBC Mode ....................... 26 3.10.2 Public Key Algorithms .............................. 26 3.10.2.1 Multi-Precision Integers .................. 27 3.10.3 Hash Functions ..................................... 27 3.10.4 MAC Algorithms ..................................... 27 3.10.5 Compression Algorithms ............................. 28 3.11 SILC Public Key .......................................... 29 3.12 SILC Version Detection ................................... 31 3.13 Backup Routers ........................................... 31 3.13.1 Switching to Backup Router ......................... 33 3.13.2 Resuming Primary Router ............................ 34 Riikonen [Page 2] Internet Draft 29 July 2003 4 SILC Procedures ............................................... 36 4.1 Creating Client Connection ................................ 37 4.2 Creating Server Connection ................................ 38 4.2.1 Announcing Clients, Channels and Servers ............ 39 4.3 Joining to a Channel ...................................... 40 4.4 Channel Key Generation .................................... 41 4.5 Private Message Sending and Reception ..................... 42 4.6 Private Message Key Generation ............................ 42 4.7 Channel Message Sending and Reception ..................... 43 4.8 Session Key Regeneration .................................. 44 4.9 Command Sending and Reception ............................. 44 4.10 Closing Connection ....................................... 45 4.11 Detaching and Resuming a Session ......................... 46 5 Security Considerations ....................................... 47 6 References .................................................... 48 7 Author's Address .............................................. 50 8 Full Copyright Statement ...................................... 50 List of Figures Figure 1: SILC Network Topology Figure 2: Communication Inside cell Figure 3: Communication Between Cells Figure 4: Router Connections Figure 5: SILC Public Key Figure 6: Counter Block 1. Introduction This document describes a Secure Internet Live Conferencing (SILC) protocol which provides secure conferencing services over insecure network channel. SILC can be used as a secure conferencing service that provides rich conferencing features. Some of the SILC features are found in traditional chat protocols such as IRC [IRC] but many of the SILC features can also be found in Instant Message (IM) style protocols. SILC combines features from both of these chat protocol styles, and can be implemented as either IRC-like system or IM-like system. Some of the more advanced and secure features of the protocol are new to all conferencing protocols. SILC also supports multimedia messages and can also be implemented as a video and audio conferencing system. Strong cryptographic methods are used to protect SILC packets inside the SILC network. Three other specifications relates very closely to this memo; SILC Packet Protocol [SILC2], SILC Key Exchange and Authentication Protocols [SILC3] and SILC Commands [SILC4]. Riikonen [Page 3] Internet Draft 29 July 2003 The protocol uses extensively packets as conferencing protocol requires message and command sending. The SILC Packet Protocol is described in [SILC2] and should be read to fully comprehend this document and protocol. [SILC2] also describes the packet encryption and decryption in detail. The SILC Packet Protocol provides secured and authenticated packets, and the protocol is designed to be compact. This makes SILC also suitable in environment of low bandwidth requirements such as mobile networks. All packet payloads in SILC can be also compressed. The security of SILC protocol sessions are based on strong and secure key exchange protocol. The SILC Key Exchange protocol is described in [SILC3] along with connection authentication protocol and should be read to fully comprehend this document and protocol. The SILC protocol has been developed to work on TCP/IP network protocol, although it could be made to work on other network protocols with only minor changes. However, it is recommended that TCP/IP protocol is used under SILC protocol. Typical implementation would be made in client-server model. 1.1 Requirements Terminology The keywords MUST, MUST NOT, REQUIRED, SHOULD, SHOULD NOT, RECOMMENDED, MAY, and OPTIONAL, when they appear in this document, are to be interpreted as described in [RFC2119]. 2. SILC Concepts This section describes various SILC protocol concepts that forms the actual protocol, and in the end, the actual SILC network. The mission of the protocol is to deliver messages from clients to other clients through routers and servers in secure manner. The messages may also be delivered from one client to many clients forming a group, also known as a channel. This section does not focus to security issues. Instead, basic network concepts are introduced to make the topology of the SILC network clear. 2.1 SILC Network Topology SILC network forms a ring as opposed to tree style network topology that conferencing protocols usually have. The network has a cells which are constructed from a router and zero or more servers. The servers are Riikonen [Page 4] Internet Draft 29 July 2003 connected to the router in a star like network topology. Routers in the network are connected to each other forming a ring. The rationale for this is to have servers that can perform specific kind of tasks what other servers cannot perform. This leads to two kinds of servers; normal SILC servers and SILC router servers. A difference between normal server and router server is that routers knows all global information and keep the global network state up to date. They also do the actual routing of the messages to the correct receiver between other cells. Normal servers knows only local information and receive global information only when it is needed. They do not need to keep the global network state up to date. This makes the network faster and scalable as there are less servers that needs to maintain global network state. This, on the other hand, leads into a cellular like network, where routers are in the center of the cell and servers are connected to the router. The following diagram represents SILC network topology. ---- ---- ---- ---- ---- ---- | S8 | S5 | S4 | | S7 | S5 | S6 | ----- ---- ----- ----- ---- ----- | S7 | S/R1 | S2 | --- | S8 | S/R2 | S4 | ---- ------ ---- ---- ------ ---- | S6 | S3 | S1 | | S1 | S3 | S2 | ---- ---- ---- ---- ---- ---- ---- ---- | S3 | S1 | Cell 1. \ Cell 2. | \____ ----- ----- | | | S4 | S/R4 | ---- ---- ---- ---- ---- ---- ---- ------ | S7 | S4 | S2 | | S1 | S3 | S2 | | S2 | S5 | ----- ---- ----- ----- ---- ----- ---- ---- | S6 | S/R3 | S1 | --- | S4 | S/R5 | S5 | ____/ Cell 4. ---- ------ ---- ---- ------ ---- | S8 | S5 | S3 | | S6 | S7 | S8 | ... etc ... ---- ---- ---- ---- ---- ---- Cell 3. Cell 5. Figure 1: SILC Network Topology A cell is formed when a server or servers connect to one router. In SILC network normal server cannot directly connect to other normal server. Normal server may only connect to SILC router which then routes the messages to the other servers in the cell. Router servers on the other hand may connect to other routers to form the actual SILC network, as seen in above figure. However, router is also able to act Riikonen [Page 5] Internet Draft 29 July 2003 as normal SILC server; clients may connect to it the same way as to normal SILC server. Normal server also cannot have active connections to more than one router. Normal server cannot be connected to two different cells. Router servers, on the other hand, may have as many router to router connections as needed. Other direct routes between other routers is also possible in addition of the mandatory ring connections. This leads into a hybrid ring-mesh network topology. There are many issues in this network topology that needs to be careful about. Issues like routing, the size of the cells, the number of the routers in the SILC network and the capacity requirements of the routers. These issues should be discussed in the Internet Community and additional documents on the issue may be written. 2.2 Communication Inside a Cell It is always guaranteed that inside a cell message is delivered to the recipient with at most two server hops. A client which is connected to server in the cell and is talking on channel to other client connected to other server in the same cell, will have its messages delivered from its local server first to the router of the cell, and from the router to the other server in the cell. The following diagram represents this scenario: 1 --- S1 S4 --- 5 S/R 2 -- S2 S3 / | 4 3 Figure 2: Communication Inside cell Example: Client 1. connected to Server 1. send message to Client 4. connected to Server 2. travels from Server 1. first to Router which routes the message to Server 2. which then sends it to the Client 4. All the other servers in the cell will not see the routed message. If the client is connected directly to the router, as router is also normal SILC server, the messages inside the cell are always delivered only with one server hop. If clients communicating with each other are connected to the same server, no router interaction is needed. This is the optimal Riikonen [Page 6] Internet Draft 29 July 2003 situation of message delivery in the SILC network. 2.3 Communication in the Network If the message is destined to client that does not belong to local cell the message is routed to the router server to which the destination client belongs, if the local router is connected to destination router. If there is no direct connection to the destination router, the local router routes the message to its primary route. The following diagram represents message sending between cells. 1 --- S1 S4 --- 5 S2 --- 1 S/R - - - - - - - - S/R 2 -- S2 S3 S1 / | \ 4 3 2 Cell 1. Cell 2. Figure 3: Communication Between Cells Example: Client 5. connected to Server 4. in Cell 1. sends message to Client 2. connected to Server 1. in Cell 2. travels from Server 4. to Router which routes the message to Router in Cell 2, which then routes the message to Server 1. All the other servers and routers in the network will not see the routed message. The optimal case of message delivery from the client point of view is when clients are connected directly to the routers and the messages are delivered from one router to the other. 2.4 Channel Communication Messages may be sent to group of clients as well. Sending messages to many clients works the same way as sending messages point to point, from message delivery point of view. Security issues are another matter which are not discussed in this section. Router server handles the message routing to multiple recipients. If any recipient is not in the same cell as the sender the messages are Riikonen [Page 7] Internet Draft 29 July 2003 routed further. Server distributes the channel message to its local clients which are joined to the channel. Router also distributes the message to its local clients on the channel. 2.5 Router Connections Router connections play very important role in making the SILC like network topology to work. For example, sending broadcast packets in SILC network require special connections between routers; routers must be connected in a specific way. Every router has their primary route which is a connection to another router in the network. Unless there is only two routers in the network must not routers use each other as their primary routes. The router connections in the network must form a ring. Example with three routers in the network: S/R1 - < - < - < - < - < - < - S/R2 \ / v ^ \ - > - > - S/R3 - > - > - / Figure 4: Router Connections Example: Network with three routers. Router 1. uses Router 2. as its primary router. Router 2. uses Router 3. as its primary router, and Router 3. uses Router 1. as its primary router. There may be other direct connections between the routers but they must not be used as primary routes. The above example is applicable to any amount of routers in the network except for two routers. If there are only two routers in the network both routers must be able to handle situation where they use each other as their primary routes. The issue of router connections are very important especially with SILC broadcast packets. Usually all router wide information in the network is distributed by SILC broadcast packets. This sort of ring network, with ability to have other direct routes in the network can cause interesting routing problems. The [SILC2] discusses the routing of packets in this sort of network in more detail. Riikonen [Page 8] Internet Draft 29 July 2003 3. SILC Specification This section describes the SILC protocol. However, [SILC2] and [SILC3] describes other important protocols that are part of this SILC specification and must be read. 3.1 Client A client is a piece of software connecting to SILC server. SILC client cannot be SILC server. Purpose of clients is to provide the user interface of the SILC services for end user. Clients are distinguished from other clients by unique Client ID. Client ID is a 128 bit ID that is used in the communication in the SILC network. The client ID is based on the user's IP address and nickname. User use logical nicknames in communication which are then mapped to the corresponding Client ID. Client IDs are low level identifications and should not be seen by the end user. Clients provide other information about the end user as well. Information such as the nickname of the user, username and the host name of the end user and user's real name. See section 3.2 Server for information of the requirements of keeping this information. The nickname selected by the user is not unique in the SILC network. There can be 2^8 same nicknames for one IP address. As for comparison to IRC [IRC] where nicknames are unique this is a fundamental difference between SILC and IRC. This typically causes the server names or client's host names to be used along with the nicknames on user interface to identify specific users when sending messages. This feature of SILC makes IRC style nickname-wars obsolete as no one owns their nickname; there can always be someone else with the same nickname. Also, any kind of nickname registering service becomes obsolete. The maximum length of nickname is 128 bytes. 3.1.1 Client ID Client ID is used to identify users in the SILC network. The Client ID is unique to the extent that there can be 2^128 different Client IDs, and IDs based on IPv6 addresses extends this to 2^224 different Client IDs. Collisions are not expected to happen. The Client ID is defined as follows. 128 bit Client ID based on IPv4 addresses: 32 bit Server ID IP address (bits 1-32) 8 bit Random number or counter Riikonen [Page 9] Internet Draft 29 July 2003 88 bit Truncated MD5 hash value of the nickname 224 bit Client ID based on IPv6 addresses: 128 bit Server ID IP address (bits 1-128) 8 bit Random number or counter 88 bit Truncated MD5 hash value of the nickname o Server ID IP address - Indicates the server where this client is coming from. The IP address hence equals the server IP address where the client is connected. o Random number or counter - Random number to further randomize the Client ID. Another choice is to use a counter starting from the zero (0). This makes it possible to have 2^8 same nicknames from the same server IP address. o MD5 hash - MD5 hash value of the lowercase nickname is truncated taking 88 bits from the start of the hash value. This hash value is used to search the user's Client ID from the ID lists. Note that the nickname MUST be in lowercase format. Collisions could occur when more than 2^8 clients using same nickname from the same server IP address is connected to the SILC network. Server MUST be able to handle this situation by refusing to accept anymore of that nickname. Another possible collision may happen with the truncated hash value of the nickname. It could be possible to have same truncated hash value for two different nicknames. However, this is not expected to happen nor cause any serious problems if it would occur. Nicknames are usually logical and it is unlikely to have two distinct logical nicknames produce same truncated hash value. 3.2 Server Servers are the most important parts of the SILC network. They form the basis of the SILC, providing a point to which clients may connect to. There are two kinds of servers in SILC; normal servers and router servers. This section focus on the normal server and router server is described in the section 3.3 Router. Normal servers MUST NOT directly connect to other normal server. Normal servers may only directly connect to router server. If the message sent by the client is destined outside the local server it is always sent to Riikonen [Page 10] Internet Draft 29 July 2003 the router server for further routing. Server may only have one active connection to router on same port. Normal server MUST NOT connect to other cell's router except in situations where its cell's router is unavailable. 3.2.1 Server's Local ID List Normal server keeps various information about the clients and their end users connected to it. Every normal server MUST keep list of all locally connected clients, Client IDs, nicknames, usernames and host names and user's real name. Normal servers only keeps local information and it does not keep any global information. Hence, normal servers knows only about their locally connected clients. This makes servers efficient as they do not have to worry about global clients. Server is also responsible of creating the Client IDs for their clients. Normal server also keeps information about locally created channels and their Channel IDs. Hence, local list for normal server includes: server list - Router connection o Server name o Server IP address o Server ID o Sending key o Receiving key o Public key client list - All clients in server o Nickname o Username@host o Real name o Client ID o Sending key o Receiving key o Public key channel list - All channels in server o Channel name o Channel ID o Client IDs on channel o Client ID modes on channel o Channel key Riikonen [Page 11] Internet Draft 29 July 2003 3.2.2 Server ID Servers are distinguished from other servers by unique 64 bit Server ID (for IPv4) or 160 bit Server ID (for IPv6). The Server ID is used in the SILC to route messages to correct servers. Server IDs also provide information for Client IDs, see section 3.1.1 Client ID. Server ID is defined as follows. 64 bit Server ID based on IPv4 addresses: 32 bit IP address of the server 16 bit Port 16 bit Random number 160 bit Server ID based on IPv6 addresses: 128 bit IP address of the server 16 bit Port 16 bit Random number o IP address of the server - This is the real IP address of the server. o Port - This is the port the server is bound to. o Random number - This is used to further randomize the Server ID. Collisions are not expected to happen in any conditions. The Server ID is always created by the server itself and server is responsible of distributing it to the router. 3.2.3 SILC Server Ports The following ports has been assigned by IANA for the SILC protocol: silc 706/tcp SILC silc 706/udp SILC If there are needs to create new SILC networks in the future the port numbers must be officially assigned by the IANA. Server on network above privileged ports (>1023) SHOULD NOT be trusted as they could have been set up by untrusted party. Riikonen [Page 12] Internet Draft 29 July 2003 3.3 Router Router server in SILC network is responsible for keeping the cell together and routing messages to other servers and to other routers. Router server is also a normal server thus clients may connect to it as it would be just normal SILC server. However, router servers has a lot of important tasks that normal servers do not have. Router server knows everything and keeps the global state. They know all clients currently on SILC, all servers and routers and all channels in SILC. Routers are the only servers in SILC that care about global information and keeping them up to date at all time. 3.3.1 Router's Local ID List Router server as well MUST keep local list of connected clients and locally created channels. However, this list is extended to include all the informations of the entire cell, not just the server itself as for normal servers. However, on router this list is a lot smaller since routers do not need to keep information about user's nickname, username and host name and real name since these are not needed by the router. The router keeps only information that it needs. Hence, local list for router includes: server list - All servers in the cell o Server name o Server ID o Router's Server ID o Sending key o Receiving key client list - All clients in the cell o Client ID channel list - All channels in the cell o Channel ID o Client IDs on channel o Client ID modes on channel o Channel key Note that locally connected clients and other information include all the same information as defined in section section 3.2.1 Server's Local ID List. Router MAY also cache same detailed information for other clients Riikonen [Page 13] Internet Draft 29 July 2003 if needed. 3.3.2 Router's Global ID List Router server MUST also keep global list. Normal servers do not have global list as they know only about local information. Global list includes all the clients on SILC, their Client IDs, all created channels and their Channel IDs and all servers and routers on SILC and their Server IDs. That is said, global list is for global information and the list must not include the local information already on the router's local list. Note that the global list does not include information like nicknames, usernames and host names or user's real names. Router does not need to keep these informations as they are not needed by the router. This information is available from the client's server which maybe queried when needed. Hence, global list includes: server list - All servers in SILC o Server name o Server ID o Router's Server ID client list - All clients in SILC o Client ID channel list - All channels in SILC o Channel ID o Client IDs on channel o Client ID modes on channel 3.3.3 Router's Server ID Router's Server ID is equivalent to normal Server ID. As routers are normal servers same types of IDs applies for routers as well. See section 3.2.2 Server ID. 3.4 Channels A channel is a named group of one or more clients which will all receive messages addressed to that channel. The channel is created when first client requests JOIN command to the channel, and the channel ceases to Riikonen [Page 14] Internet Draft 29 July 2003 exist when the last client has left it. When channel exists, any client can reference it using the Channel ID of the channel. If the channel has a founder mode set and last client leaves the channel the channel does not cease to exist. The founder mode can be used to make permanent channels in the network. The founder of the channel can regain the channel founder privileges on the channel later when he joins the channel. Channel names are unique although the real uniqueness comes from 64 bit Channel ID. However, channel names are still unique and no two global channels with same name may exist. The channel name is a string of maximum length of 256 bytes. Channel names MUST NOT contain any whitespaces (` '), any non-printable ASCII characters, commas (`,') and wildcard characters. Channels can have operators that can administrate the channel and operate all of its modes. The following operators on channel exist on the SILC network. o Channel founder - When channel is created the joining client becomes channel founder. Channel founder is channel operator with some more privileges. Basically, channel founder can fully operate the channel and all of its modes. The privileges are limited only to the particular channel. There can be only one channel founder per channel. Channel founder supersedes channel operator's privileges. Channel founder privileges cannot be removed by any other operator on channel. When channel founder leaves the channel there is no channel founder on the channel. However, it is possible to set a mode for the channel which allows the original channel founder to regain the founder privileges even after leaving the channel. Channel founder also cannot be removed by force from the channel. o Channel operator - When client joins to channel that has not existed previously it will become automatically channel operator (and channel founder discussed above). Channel operator is able to administrate the channel, set some modes on channel, remove a badly behaving client from the channel and promote other clients to become channel operator. The privileges are limited only to the particular channel. Normal channel user may be promoted (opped) to channel operator gaining channel operator privileges. Channel founder or other channel operator may also demote (deop) channel operator to normal channel user. Riikonen [Page 15] Internet Draft 29 July 2003 3.4.1 Channel ID Channels are distinguished from other channels by unique Channel ID. The Channel ID is a 64 bit ID (for IPv4) or 160 bit ID (for IPv6), and collisions are not expected to happen in any conditions. Channel names are just for logical use of channels. The Channel ID is created by the server where the channel is created. The Channel ID is defined as follows. 64 bit Channel ID based on IPv4 addresses: 32 bit Router's Server ID IP address (bits 1-32) 16 bit Router's Server ID port (bits 33-48) 16 bit Random number or counter 160 bit Channel ID based on IPv6 addresses: 128 bit Router's Server ID IP address (bits 1-128) 16 bit Router's Server ID port (bits 129-144) 16 bit Random number or counter o Router's Server ID IP address - Indicates the IP address of the router of the cell where this channel is created. This is taken from the router's Server ID. This way SILC router knows where this channel resides in the SILC network. o Router's Server ID port - Indicates the port of the channel on the server. This is taken from the router's Server ID. o Random number or counter - To further randomize the Channel ID. Another choice is to use a counter starting from zero (0). This makes sure that there are no collisions. This also means that in a cell there can be 2^16 different channels. 3.5 Operators Operators are normal users with extra privileges to their server or router. Usually these people are SILC server and router administrators that take care of their own server and clients on them. The purpose of operators is to administrate the SILC server or router. However, even an operator with highest privileges is not able to enter invite-only channels, to gain access to the contents of encrypted and authenticated packets traveling in the SILC network or to gain channel operator privileges on public channels without being promoted. They have the same privileges as any normal user except they are able to administrate their server or router. Riikonen [Page 16] Internet Draft 29 July 2003 3.6 SILC Commands Commands are very important part on SILC network especially for client which uses commands to operate on the SILC network. Commands are used to set nickname, join to channel, change modes and many other things. Client usually sends the commands and server replies by sending a reply packet to the command. Server MAY also send commands usually to serve the original client's request. Usually server cannot send commands to clients, however there MAY be commands that allow the server to send commands to client. By default servers MAY send commands only to other servers and routers. Note that the command reply is usually sent only after client has sent the command request but server is allowed to send command reply packet to client even if client has not requested the command. Client MAY choose to ignore the command reply. It is expected that some of the commands may be misused by clients resulting various problems on the server side. Every implementation SHOULD assure that commands may not be executed more than once, say, in two (2) seconds. However, to keep response rate up, allowing for example five (5) commands before limiting is allowed. It is RECOMMENDED that commands such as SILC_COMMAND_NICK, SILC_COMMAND_JOIN, SILC_COMMAND_LEAVE and SILC_COMMAND_KILL SHOULD be limited in all cases as they require heavy operations. This should be sufficient to prevent the misuse of commands. SILC commands are described in [SILC4]. 3.7 SILC Packets Packets are naturally the most important part of the protocol and the packets are what actually makes the protocol. Packets in SILC network are always encrypted using, usually the shared secret session key or some other key, for example, channel key, when encrypting channel messages. It is not possible to send a packet in SILC network without encryption. The SILC Packet Protocol is a wide protocol and is described in [SILC2]. This document does not define or describe details of SILC packets. 3.8 Packet Encryption All packets passed in SILC network MUST be encrypted. This section gives generic description of how packets must be encrypted in the SILC network. The detailed description of the actual encryption process Riikonen [Page 17] Internet Draft 29 July 2003 of the packets are described in [SILC2]. Client and its server shares secret symmetric session key which is established by the SILC Key Exchange Protocol, described in [SILC3]. Every packet sent from client to server, with exception of packets for channels, are encrypted with this session key. Channels have a channel key that are shared by every client on the channel. However, the channel keys are cell specific thus one cell does not know the channel key of the other cell, even if that key is for same channel. Channel key is also known by the routers and all servers that have clients on the channel. However, channels MAY have channel private keys that are entirely local setting for the client. All clients on the channel MUST know the channel private key beforehand to be able to talk on the channel. In this case, no server or router knows the key for the channel. Server shares secret symmetric session key with router which is established by the SILC Key Exchange Protocol. Every packet passed from server to router, with exception of packets for channels, are encrypted with the shared session key. Same way, router server shares secret symmetric key with its primary router. However, every packet passed from router to other router, including packets for channels, are encrypted with the shared session key. Every router connection MUST have their own session keys. 3.8.1 Determination of the Source and the Destination The source and the destination of the packet needs to be determined to be able to route the packets to correct receiver. This information is available in the SILC Packet Header which is included in all packets sent in SILC network. The SILC Packet Header is described in [SILC2]. The header MUST be encrypted with the session key of whom is the next receiver of the packet along the route. The receiver of the packet, for example a router along the route, is able to determine the sender and the destination of the packet by decrypting the SILC Packet Header and checking the IDs attached to the header. The IDs in the header will tell to where the packet needs to be sent and where it is coming from. The header in the packet MUST NOT change during the routing of the packet. The original sender, for example client, assembles the packet and the packet header and server or router between the sender and the receiver MUST NOT change the packet header. Note however, that some packets such as commands may be resent by a server to serve the client's original command. In this case the command packet sent by the server includes the server's IDs as it is a different packet. When server or router receives a packet it MUST verify that the Source ID is Riikonen [Page 18] Internet Draft 29 July 2003 valid and correct ID for that sender. Note that the packet and the packet header may be encrypted with different keys. For example, packets to channels are encrypted with the channel key, however, the header is encrypted with the session key as described above. However, the header and the packet may be encrypted with same key. This is the case, for example, with command packets. 3.8.2 Client To Client The process of message delivery and encryption from client to another client is as follows. Example: Private message from client to another client on different servers. Clients do not share private message delivery keys; normal session keys are used. o Client 1 sends encrypted packet to its server. The packet is encrypted with the session key shared between client and its server. o Server determines the destination of the packet and decrypts the packet. Server encrypts the packet with session key shared between the server and its router, and sends the packet to the router. o Router determines the destination of the packet and decrypts the packet. Router encrypts the packet with session key shared between the router and the destination server, and sends the packet to the server. o Server determines the client to which the packet is destined to and decrypts the packet. Server encrypts the packet with session key shared between the server and the destination client, and sends the packet to the client. o Client 2 decrypts the packet. Example: Private message from client to another client on different servers. Clients have established a secret shared private message delivery key with each other and that is used in the message encryption. o Client 1 sends encrypted packet to its server. The packet header is encrypted with the session key shared between the client and server, and the private message is encrypted with the private Riikonen [Page 19] Internet Draft 29 July 2003 message delivery key shared between clients. o Server determines the destination of the packet and sends the packet to the router. Header is encrypted with the session key. o Router determines the destination of the packet and sends the packet to the server. Header is encrypted with the session key. o Server determines the client to which the packet is destined to and sends the packet to the client. Header is encrypted with the session key. o Client 2 decrypts the packet with the secret shared key. If clients share secret key with each other the private message delivery is much simpler since servers and routers between the clients do not need to decrypt and re-encrypt the entire packet. The packet header however is always encrypted with session key and is decrypted and re-encrypted with the session key of next recipient. The process for clients on same server is much simpler as there is no need to send the packet to the router. The process for clients on different cells is same as above except that the packet is routed outside the cell. The router of the destination cell routes the packet to the destination same way as described above. 3.8.3 Client To Channel Process of message delivery from client on channel to all the clients on the channel. Example: Channel of four users; two on same server, other two on different cells. Client sends message to the channel. Packet header is encrypted with the session key, message data is encrypted with channel key. o Client 1 encrypts the packet with channel key and sends the packet to its server. o Server determines local clients on the channel and sends the packet to the Client on the same server. Server then sends the packet to its router for further routing. o Router determines local clients on the channel, if found sends packet to the local clients. Router determines global clients on the channel and sends the packet to its primary router or fastest route. Riikonen [Page 20] Internet Draft 29 July 2003 o (Other router(s) do the same thing and sends the packet to the server(s).) o Server determines local clients on the channel and sends the packet to the client. o All clients receiving the packet decrypts it. 3.8.4 Server To Server Server to server packet delivery and encryption is described in above examples. Router to router packet delivery is analogous to server to server. However, some packets, such as channel packets, are processed differently. These cases are described later in this document and more in detail in [SILC2]. 3.9 Key Exchange And Authentication Key exchange is done always when for example client connects to server but also when server and router, and router and another router connect to each other. The purpose of key exchange protocol is to provide secure key material to be used in the communication. The key material is used to derive various security parameters used to secure SILC packets. The SILC Key Exchange protocol is described in detail in [SILC3]. Authentication is done after key exchange protocol has been successfully completed. The purpose of authentication is to authenticate for example client connecting to the server. However, clients MAY be accepted to connect to server without explicit authentication. Servers are REQUIRED to use authentication protocol when connecting. The authentication may be based on passphrase (pre-shared secret) or public key based on digital signatures. All passphrases sent in SILC protocol MUST be UTF-8 [RFC2279] encoded. The connection authentication protocol is described in detail in [SILC3]. 3.9.1 Authentication Payload Authentication Payload is used separately from the SKE and the Connection Authentication protocols. It can be used during the session to authenticate with a remote. For example, a client can authenticate itself to a server to become server operator. In this case, Authentication Payload is used. The format of the Authentication Payload is as follows: Riikonen [Page 21] Internet Draft 29 July 2003 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Payload Length | Authentication Method | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Public Data Length | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | | ~ Public Data ~ | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Authentication Data Length | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | | ~ Authentication Data ~ | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Figure 5: Authentication Payload o Payload Length (2 bytes) - Length of the entire payload. o Authentication Method (2 bytes) - The method of the authentication. The authentication methods are defined in [SILC2] in the Connection Auth Request Payload. The NONE authentication method SHOULD NOT be used. o Public Data Length (2 bytes) - Indicates the length of the Public Data field. o Public Data (variable length) - This is defined only if the authentication method is public key. If it is any other this field MAY include random data for padding purposes. However, in this case the field MUST be ignored by the receiver. When the authentication method is public key this includes 128 to 4096 bytes of non-zero random data that is used in the signature process, described subsequently. o Authentication Data Length (2 bytes) - Indicates the length of the Authentication Data field. If zero (0) value is found in this field the payload MUST be discarded. o Authentication Data (variable length) - Authentication method dependent authentication data. Riikonen [Page 22] Internet Draft 29 July 2003 If the authentication method is passphrase-based, the Authentication Data field includes the plaintext UTF-8 encoded passphrase. It is safe to send plaintext passphrase since the entire payload is encrypted. In this case the Public Data Length is set to zero (0), but MAY also include random data for padding purposes. It is also RECOMMENDED that maximum amount of padding is applied to SILC packet when using passphrase-based authentication. This way it is not possible to approximate the length of the passphrase from the encrypted packet. If the authentication method is public key based (or certificate) the Authentication Data is computed as follows: HASH = hash(random bytes | ID | public key (or certificate)); Authentication Data = sign(HASH); The hash() and the sign() are the hash function and the public key cryptography function selected in the SKE protocol, unless otherwise stated in the context where this payload is used. The public key is SILC style public key unless certificates are used. The ID is the entity's ID (Client or Server ID) which is authenticating itself. The ID encoding is described in [SILC2]. The random bytes are non-zero random bytes of length between 128 and 4096 bytes, and will be included into the Public Data field as is. The receiver will compute the signature using the random data received in the payload, the ID associated to the connection and the public key (or certificate) received in the SKE protocol. After computing the receiver MUST verify the signature. Also in case of public key authentication this payload is encrypted. 3.10 Algorithms This section defines all the allowed algorithms that can be used in the SILC protocol. This includes mandatory cipher, mandatory public key algorithm and MAC algorithms. 3.10.1 Ciphers Cipher is the encryption algorithm that is used to protect the data in the SILC packets. See [SILC2] for the actual encryption process and definition of how it must be done. SILC has a mandatory algorithm that must be supported in order to be compliant with this protocol. The following ciphers are defined in SILC protocol: aes-256-cbc AES in CBC mode, 256 bit key (REQUIRED) Riikonen [Page 23] Internet Draft 29 July 2003 aes-256-ctr AES in CTR mode, 256 bit key (RECOMMENDED) aes-256-rcbc AES in randomized CBC mode, 256 bit key (OPTIONAL) aes-192- AES in mode, 192 bit key (OPTIONAL) aes-128- AES in mode, 128 bit key (RECOMMENDED) twofish-256- Twofish in mode, 256 bit key (OPTIONAL) twofish-192- Twofish in mode, 192 bit key (OPTIONAL) twofish-128- Twofish in mode, 128 bit key (OPTIONAL) cast-256- CAST-256 in mode, 256 bit key (OPTIONAL) cast-192- CAST-256 in mode, 192 bit key (OPTIONAL) cast-128- CAST-256 in mode, 128 bit key (OPTIONAL) serpent-- Serpent in mode, bit key (OPTIONAL) rc6-- RC6 in mode, bit key (OPTIONAL) mars-- MARS in mode, bit key (OPTIONAL) none No encryption (OPTIONAL) The is either "cbc", "ctr" or "rcbc". Other encryption modes MAY be defined to be used in SILC using the same name format. The is either 256, 192 or 128 bit key length. Also, additional ciphers MAY be defined to be used in SILC by using the same name format as above. Algorithm "none" does not perform any encryption process at all and thus is not recommended to be used. It is recommended that no client or server implementation would accept none algorithm except in special debugging mode. 3.10.1.1 CBC Mode The "cbc" encryption mode is CBC mode with inter-packet chaining. This means that the Initialization Vector (IV) for the next encryption block is the previous ciphertext block. The very first IV MUST be random and is generated as described in [SILC3]. 3.10.1.2 CTR Mode The "ctr" encryption mode is Counter Mode (CTR). The CTR mode in SILC is stateful in encryption and decryption. Both sender and receiver maintain the counter for the CTR mode and thus can precompute the key stream for encryption and decryption. By default, CTR mode does not require plaintext padding, however implementations MAY apply padding to the packets. If the last key block is larger than the last plaintext block the resulted value is truncated to the size of the plaintext block and the most significant bits are used. When sending authentication data inside packets the maximum amount of padding SHOULD be applied with CTR mode as well. In CTR mode only the encryption operation of the cipher is used. The Riikonen [Page 24] Internet Draft 29 July 2003 decryption operation is not needed since both encryption and decryption process is simple XOR with the plaintext block and the key stream block. The counter block is used to create the key for the CTR mode. When SILC specifications refer to Initialization Vector (IV) in general cases, in case of CTR mode it refers to the counter block. The format of the 128 bit counter block is as follows: 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Truncated HASH from SKE | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Sending/Receiving IV from SKE | | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Block Counter | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Figure 6: Counter Block o Truncated HASH from SKE (4 bytes) - This value is the first 4 bytes from the HASH value that was computed as a result of SKE protocol. This acts as session identifier and each rekey MUST produce a new HASH value. o Sending/Receiving IV from SKE (8 bytes) - This value is the first 8 bytes from the Sending IV or Receiving IV generated in the SKE protocol. When this mode is used to encrypt sending traffic the Sending IV is used, when used to decrypt receiving traffic the Receiving IV is used. This assures that two parties of the protocol use different IV for sending traffic. Each rekey MUST produce a new value. o Block Counter (4 bytes) - This is the counter value for the counter block and is MSB ordered number starting from one (1) value for first block and incrementing for subsequent blocks. The same value MUST NOT be used twice. The rekey MUST be performed before this counter value wraps. CTR mode MUST NOT be used with "none" MAC. Implementations also MUST assure that the same counter block is not used to encrypt more than one block. Also, the key material used with CTR mode MUST be fresh key material. Static keys (pre-shared keys) MUST NOT be used with CTR mode. For this reason using CTR mode to encrypt for example channel messages or private messages with a pre-shared key is inappropriate. For private messages, the Key Agreement could be performed to produce fresh key material. Riikonen [Page 25] Internet Draft 29 July 2003 If the IV Included flag was negotiated in SKE, implementations SHOULD still use the same counter block format as defined above. However, implementations are RECOMMENDED to replace the Truncated HASH field with a 32 bit random value for each IV (counter block) per encrypted SILC packet. Also note, that in this case the decryption process is not stateful and receiver cannot precompute the key stream. 3.10.1.3 Randomized CBC Mode The "rcbc" encryption mode is CBC mode with randomized IV. This means that each IV for each packet MUST be chosen randomly. When encrypting more than one block the normal inter-packet chaining is used, but for the first block new random IV is selected in each packet. In this mode the IV is appended at the end of the last ciphertext block and thus delivered to the recipient. This mode increases the ciphertext size by one ciphertext block. Note also that some data payloads in SILC are capable of delivering the IV to the recipient. When explicitly encrypting these payloads with randomized CBC the IV MUST NOT be appended at the end of the ciphertext, but is placed at the specified location in the payload. However, Message Payload for example has the IV at the location which is equivalent to placing it after the last ciphertext block. When using CBC mode with such payloads it is actually equivalent to using randomized CBC since the IV is selected in random and included in the ciphertext. 3.10.2 Public Key Algorithms Public keys are used in SILC to authenticate entities in SILC network and to perform other tasks related to public key cryptography. The public keys are also used in the SILC Key Exchange protocol [SILC3]. The following public key algorithms are defined in SILC protocol: rsa RSA (REQUIRED) dss DSS (OPTIONAL) DSS is described in [Menezes]. The RSA MUST be implemented according PKCS #1 [PKCS1]. The mandatory PKCS #1 implementation in SILC MUST be compliant to either PKCS #1 version 1.5 or newer with the following notes: The signature encoding is always in same format as the encryption encoding regardless of the PKCS #1 version. The signature with appendix (with hash algorithm OID in the data) MUST NOT be used in the SILC. The rationale for this is that there is no binding between the PKCS #1 OIDs and the hash algorithms used in the SILC protocol. Hence, the encoding is always in PKCS #1 version 1.5 format. Riikonen [Page 26] Internet Draft 29 July 2003 Additional public key algorithms MAY be defined to be used in SILC. When signatures are computed in SILC the computing of the signature is represented as sign(). The signature computing procedure is dependent of the public key algorithm, and the public key or certificate encoding. When using SILC public key the signature is computed as described in previous paragraph for RSA and DSS keys. If the hash function is not specified separately for signing process SHA-1 MUST be used. When using SSH2 public keys the signature is computed as described in [SSH-TRANS]. When using X.509 version 3 certificates the signature is computed as described in [PKCS7]. When using OpenPGP certificates the signature is computed as described in [PGP]. 3.10.2.1 Multi-Precision Integers Multi-Precision (MP) integers in SILC are encoded and decoded as defined in PKCS #1 [PKCS1]. MP integers are unsigned, encoded with desired octet length. This means that if the octet length is more than the actual length of the integer one or more leading zero octets will appear at the start of the encoding. The actual length of the integer is the bit size of the integer not counting any leading zero bits. 3.10.3 Hash Functions Hash functions are used as part of MAC algorithms defined in the next section. They are also used in the SILC Key Exchange protocol defined in the [SILC3]. The following Hash algorithm are defined in SILC protocol: sha1 SHA-1, length = 20 (REQUIRED) md5 MD5, length = 16 (RECOMMENDED) 3.10.4 MAC Algorithms Data integrity is protected by computing a message authentication code (MAC) of the packet data. See [SILC2] for details how to compute the MAC for a packet. The following MAC algorithms are defined in SILC protocol: hmac-sha1-96 HMAC-SHA1, length = 12 bytes (REQUIRED) hmac-md5-96 HMAC-MD5, length = 12 bytes (OPTIONAL) hmac-sha1 HMAC-SHA1, length = 20 bytes (OPTIONAL) Riikonen [Page 27] Internet Draft 29 July 2003 hmac-md5 HMAC-MD5, length = 16 bytes (OPTIONAL) none No MAC (OPTIONAL) The "none" MAC is not recommended to be used as the packet is not authenticated when MAC is not computed. It is recommended that no client or server would accept none MAC except in special debugging mode. The HMAC algorithm is described in [HMAC]. The hash algorithms used in HMACs, the SHA-1 is described in [RFC3174] and MD5 is described in [RFC1321]. Additional MAC algorithms MAY be defined to be used in SILC. 3.10.5 Compression Algorithms SILC protocol supports compression that may be applied to unencrypted data. It is recommended to use compression on slow links as it may significantly speed up the data transmission. By default, SILC does not use compression which is the mode that must be supported by all SILC implementations. The following compression algorithms are defined: none No compression (REQUIRED) zlib GNU ZLIB (LZ77) compression (OPTIONAL) Additional compression algorithms MAY be defined to be used in SILC. 3.11 SILC Public Key This section defines the type and format of the SILC public key. All implementations MUST support this public key type. See [SILC3] for other optional public key and certificate types allowed in the SILC protocol. Public keys in SILC may be used to authenticate entities and to perform other tasks related to public key cryptography. The format of the SILC Public Key is as follows: Riikonen [Page 28] Internet Draft 29 July 2003 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Public Key Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Algorithm Name Length | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | | ~ Algorithm Name ~ | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Identifier Length | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | | ~ Identifier ~ | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | ~ Public Data ~ | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Figure 5: SILC Public Key o Public Key Length (4 bytes) - Indicates the full length of the SILC Public Key, not including this field. o Algorithm Name Length (2 bytes) - Indicates the length of the Algorithm Length field, not including this field. o Algorithm name (variable length) - Indicates the name of the public key algorithm that the key is. See the section 3.10.2 Public Key Algorithms for defined names. o Identifier Length (2 bytes) - Indicates the length of the Identifier field, not including this field. o Identifier (variable length) - Indicates the identifier of the public key. This data can be used to identify the owner of the key. The identifier is of the following format: UN User name HN Host name or IP address RN Real name E EMail address O Organization Riikonen [Page 29] Internet Draft 29 July 2003 C Country Examples of an identifier: `UN=priikone, HN=poseidon.pspt.fi, E=priikone@poseidon.pspt.fi' `UN=sam, HN=dummy.fi, RN=Sammy Sam, O=Company XYZ, C=Finland' At least user name (UN) and host name (HN) MUST be provided as identifier. The fields are separated by commas (`,'). If comma is in the identifier string it must be escaped as `\,', for example, `O=Company XYZ\, Inc.'. Other characters that require escaping are listed in [RFC2253] and are to be escaped as defined therein. o Public Data (variable length) - Includes the actual public data of the public key. The format of this field for RSA algorithm is as follows: 4 bytes Length of e variable length e 4 bytes Length of n variable length n The format of this field for DSS algorithm is as follows: 4 bytes Length of p variable length p 4 bytes Length of q variable length q 4 bytes Length of g variable length g 4 bytes Length of y variable length y The variable length fields are multiple precession integers encoded as strings in both examples. Other algorithms must define their own type of this field if they are used. All fields in the public key are in MSB (most significant byte first) order. All strings in the public key MUST be UTF-8 encoded. Riikonen [Page 30] Internet Draft 29 July 2003 If an external protocol needs to refer to SILC Public Key by name, the names "silc-rsa" and "silc-dss" for SILC Public Key based on RSA algorithm and SILC Public Key based on DSS algorithm, respectively, are to be used. However, this SILC specification does not use these names directly, and they are defined here for external protocols (protocols that may like to use SILC Public Key). 3.12 SILC Version Detection The version detection of both client and server is performed at the connection phase while executing the SILC Key Exchange protocol. The version identifier is exchanged between initiator and responder. The version identifier is of the following format: SILC-- The version strings are of the following format: protocol version = . software version = [.[.]] Protocol version MUST provide both major and minor version. Currently implementations MUST set the protocol version and accept at least the protocol version as SILC-1.2-. If new protocol version causes incompatibilities with older version the version number MUST be incremented. The is incremented if new protocol version is fully incompatible. Software version MAY provide major, minor and build (vendor) version. The software version MAY be freely set and accepted. The version string MUST consist of printable US-ASCII characters. Thus, the version strings could be, for example: SILC-1.1-2.0.2 SILC-1.0-1.2 SILC-1.2-1.0.VendorXYZ SILC-1.2-2.4.5 Vendor Limited 3.13 Backup Routers Backup routers may exist in the cell in addition to the primary router. However, they must not be active routers or act as routers in the cell. Only one router may be acting as primary router in the cell. In the case of failure of the primary router one of the backup routers becomes active. The purpose of backup routers are in case of failure of the primary router Riikonen [Page 31] Internet Draft 29 July 2003 to maintain working connections inside the cell and outside the cell and to avoid netsplits. Backup routers are normal servers in the cell that are prepared to take over the tasks of the primary router if needed. They need to have at least one direct and active connection to the primary router of the cell. This communication channel is used to send the router information to the backup router. When the backup router connects to the primary router of the cell it MUST present itself as router server in the Connection Authentication protocol, even though it is normal server as long as the primary router is available. Reason for this is that the configuration needed in the responder end requires usually router connection level configuration. The responder, however must understand and treat the connection as normal server (except when feeding router level data to the backup router). Backup router must know everything that the primary router knows to be able to take over the tasks of the primary router. It is the primary router's responsibility to feed the data to the backup router. If the backup router does not know all the data in the case of failure some connections may be lost. The primary router of the cell must consider the backup router being an actual router server when it feeds the data to it. In addition to having direct connection to the primary router of the cell, the backup router must also have connection to the same router to which the primary router of the cell is connected. However, it must not be the active router connection meaning that the backup router must not use that channel as its primary route and it must not notify the router about having connected servers, channels and clients behind it. It merely connects to the router. This sort of connection is later referred to as being a passive connection. Some keepalive actions may be needed by the router to keep the connection alive. It is required that other normal servers have passive connections to the backup router(s) in the cell. Some keepalive actions may be needed by the server to keep the connection alive. After they notice the failure of the primary router they must start using the connection to the first backup router as their primary route. Also, if any other router in the network is using the cell's primary router as its own primary router, it must also have passive connection to the cell's backup router. It too is prepared to switch to use the backup router as its new primary router as soon as the original primary router becomes unresponsive. All of the parties of this protocol know which one is the backup router of the cell from their local configuration. Each of the entities must Riikonen [Page 32] Internet Draft 29 July 2003 be configured accordingly and care must be taken when configuring the backup routers, servers and other routers in the network. It must be noted that some of the channel messages and private messages may be lost during the switch to the backup router. The announcements assure that the state of the network is not lost during the switch. It is RECOMMENDED that there would be at least one backup router in the cell. It is NOT RECOMMENDED to have all servers in the cell acting as backup routers as it requires establishing several connections to several servers in the cell. Large cells can easily have several backup routers in the cell. The order of the backup routers are decided at the local configuration phase. All the parties of this protocol must be configured accordingly to understand the order of the backup routers. It is not required that the backup server is actually an active server in the cell. The backup router may be a redundant server in the cell that does not accept normal client connections at all. It may be reserved purely for the backup purposes. If also the first backup router is down as well and there is another backup router in the cell then it will start acting as the primary router as described above. 3.13.1 Switching to Backup Router When the primary router of the cell becomes unresponsive, for example by sending EOF to the connection, all the parties of this protocol MUST replace the old connection to the primary router with first configured backup router. The backup router usually needs to do local modifications to its database in order to update all the information needed to maintain working routes. The backup router must understand that clients that were originated from the primary router are now originated from some of the existing server connections and must update them accordingly. It must also remove those clients that were owned by the primary router since those connections were lost when the primary router became unresponsive. All the other parties of the protocol must also update their local database to understand that the route to the primary router will now go to the backup router. Servers connected to the backup router MUST send SILC_PACKET_RESUME_ROUTER packet with type value 21, to indicate that the server will start using the backup router as primary router. The backup router MUST NOT allow this action if it detects that primary is still up and running. If backup router knows that primary is up and running it MUST send Riikonen [Page 33] Internet Draft 29 July 2003 SILC_PACKET_FAILURE with type value 21 (4 bytes, MSB first order) back to the server. The server then MUST NOT use the backup as primary router, but must try to establish connection back to the primary router. If the action is allowed type value 21 is sent back to the server from the backup router. It is RECOMMENDED that implementations use the SILC_COMMAND_PING command to detect whether primary router is responsive. The servers connected to the backup router must then announce their clients, channels, channel users, channel user modes, channel modes, topics and other information to the backup router. This is to assure that none of the important notify packets were lost during the switch to the backup router. The backup router must check which of these announced entities it already has and distribute the new ones to the primary router. The backup router too must announce its servers, clients, channels and other information to the new primary router. The primary router of the backup router too must announce its information to the backup router. Both must process only the ones they do not know about. If any of the announced modes do not match then they are enforced in normal manner as defined in section 4.2.1 Announcing Clients, Channels and Servers. 3.13.2 Resuming Primary Router Usually the primary router is unresponsive only a short period of time and it is intended that the original router of the cell will resume its position as primary router when it comes back online. The backup router that is now acting as primary router of the cell must constantly try to connect to the original primary router of the cell. It is RECOMMENDED that it would try to reconnect in 30 second intervals to the primary router. When the connection is established to the primary router the backup resuming protocol is executed. The protocol is advanced as follows: 1. Backup router sends SILC_PACKET_RESUME_ROUTER packet with type value 1 to the primary router that came back online. The packet will indicate the primary router has been replaced by the backup router. After sending the packet the backup router will announce all of its channels, channel users, modes etc. to the primary router. If the primary knows that it has not been replaced (for example the backup itself disconnected from the primary router and thinks that it is now primary in the cell) the primary router send SILC_PACKET_FAILURE with the type value 1 (4 bytes, MSB first Riikonen [Page 34] Internet Draft 29 July 2003 order) back to the backup router. If backup receives this it MUST NOT continue with the backup resuming protocol. 2. Backup router sends SILC_PACKET_RESUME_ROUTER packet with type value 1 to its current primary router to indicate that it will resign as being primary router. Then, backup router sends the SILC_PACKET_RESUME_ROUTER packet with type value 1 to all connected servers to also indicate that it will resign as being primary router. 3. Backup router also send SILC_PACKET_RESUME_ROUTER packet with type value 1 to the router that is using the backup router currently as its primary router. 4. Any server and router that receives the SILC_PACKET_RESUME_ROUTER with type value 1 must reconnect immediately to the primary router of the cell that came back online. After they have created the connection they MUST NOT use that connection as active primary route but still route all packets to the backup router. After the connection is created they MUST send SILC_PACKET_RESUME_ROUTER with type value 2 back to the backup router. The session ID value found in the first packet MUST be set in this packet. 5. Backup router MUST wait for all packets with type value 2 before it continues with the protocol. It knows from the session ID values set in the packet when it has received all packets. The session value should be different in all packets it has sent earlier. After the packets are received the backup router sends the SILC_PACKET_RESUME_ROUTER packet with type value 3 to the primary router that came back online. This packet will indicate that the backup router is now ready to resign as being primary router. The session ID value in this packet MUST be the same as in the first packet sent to the primary router. During this time the backup router must still route all packets it is receiving from server connections. 6. The primary router receives the packet and send the packet SILC_PACKET_RESUME_ROUTER with type value 4 to all connected servers including the backup router. It also sends the packet with type value 4 to its primary router, and to the router that is using it as its primary router. The Session ID value in this packet SHOULD be zero (0). 7. Any server and router that receives the SILC_PACKET_RESUME_ROUTER packet with type value 4 must switch their primary route to the new primary router and remove the route for the backup router, since it is no longer the primary router of the cell. They must also update their local database to understand that the clients are Riikonen [Page 35] Internet Draft 29 July 2003 not originated from the backup router but from the locally connected servers. After that they MUST announce their channels, channel users, modes etc. to the primary router. They MUST NOT use the backup router connection after this and the connection is considered to be a passive connection. The implementation SHOULD be able to disable the connection without closing the actual link. After this protocol is executed the backup router is now again a normal server in the cell that has the backup link to the primary router. The primary router feeds the router specific data again to the backup router. All server connections to the backup router are considered passive connections. When the primary router of the cell comes back online and connects to its remote primary router, the remote primary router MUST send the SILC_PACKET_RESUME_ROUTER packet with type value 20 indicating that the connection is not allowed since the router has been replaced by an backup router in the cell. The session ID value in this packet SHOULD be zero (0). When the primary router receives this packet it MUST NOT use the connection as active connection but must understand that it cannot act as primary router in the cell, until the backup resuming protocol has been executed. The following type values has been defined for SILC_PACKET_RESUME_ROUTER packet: 1 SILC_SERVER_BACKUP_START 2 SILC_SERVER_BACKUP_START_CONNECTED 3 SILC_SERVER_BACKUP_START_ENDING 4 SILC_SERVER_BACKUP_START_RESUMED 20 SILC_SERVER_BACKUP_START_REPLACED 21 SILC_SERVER_BACKUP_START_USE If any other value is found in the type field the packet MUST be discarded. The SILC_PACKET_RESUME_ROUTER packet and its payload is defined in [SILC2]. 4 SILC Procedures This section describes various SILC procedures such as how the connections are created and registered, how channels are created and so on. The references [SILC2], [SILC3] and [SILC4] permeate this section's definitions. Riikonen [Page 36] Internet Draft 29 July 2003 4.1 Creating Client Connection This section describes the procedure when a client connects to SILC server. When client connects to server the server MUST perform IP address lookup and reverse IP address lookup to assure that the origin host really is who it claims to be. Client, a host, connecting to server SHOULD have both valid IP address and fully qualified domain name (FQDN). After that the client and server performs SILC Key Exchange protocol which will provide the key material used later in the communication. The key exchange protocol MUST be completed successfully before the connection registration may continue. The SILC Key Exchange protocol is described in [SILC3]. Typical server implementation would keep a list of connections that it allows to connect to the server. The implementation would check, for example, the connecting client's IP address from the connection list before the SILC Key Exchange protocol has been started. The reason for this is that if the host is not allowed to connect to the server there is no reason to perform the key exchange protocol. After successful key exchange protocol the client and server perform connection authentication protocol. The purpose of the protocol is to authenticate the client connecting to the server. Flexible implementation could also accept the client to connect to the server without explicit authentication. However, if authentication is desired for a specific client it may be based on passphrase or public key authentication. If authentication fails the connection MUST be terminated. The connection authentication protocol is described in [SILC3]. After successful key exchange and authentication protocol the client MUST register itself by sending SILC_PACKET_NEW_CLIENT packet to the server. This packet includes various information about the client that the server uses to register the client. Server registers the client and sends SILC_PACKET_NEW_ID to the client which includes the created Client ID that the client MUST start using after that. After that all SILC packets from the client MUST have the Client ID as the Source ID in the SILC Packet Header, described in [SILC2]. Client MUST also get the server's Server ID that is to be used as Destination ID in the SILC Packet Header when communicating with the server (for example when sending commands to the server). The ID may be resolved in two ways. Client can take the ID from an previously received packet from server that MUST include the ID, or to send SILC_COMMAND_INFO command and receive the Server ID as command reply. Riikonen [Page 37] Internet Draft 29 July 2003 Server MAY choose not to use the information received in the SILC_PACKET_NEW_CLIENT packet. For example, if public key or certificate were used in the authentication, server MAY use that information rather than what it received from client. This is a suitable way to get the true information about client if it is available. The nickname of client is initially set to the username sent in the SILC_PACKET_NEW_CLIENT packet. User may set the nickname to something more desirable by sending SILC_COMMAND_NICK command. However, this is not required as part of registration process. Server MUST also distribute the information about newly registered client to its router (or if the server is router, to all routers in the SILC network). More information about this in [SILC2]. Router server MUST also check whether some client in the local cell is watching for the nickname this new client has, and send the SILC_NOTIFY_TYPE_WATCH to the watcher. 4.2 Creating Server Connection This section describes the procedure when server connects to its router (or when router connects to other router, the cases are equivalent). The procedure is very much alike to when a client connects to the server thus it is not repeated here. One difference is that server MUST perform connection authentication protocol with proper authentication. A proper authentication is based on passphrase authentication or public key authentication based on digital signatures. After server and router have successfully performed the key exchange and connection authentication protocol, the server MUST register itself to the router by sending SILC_PACKET_NEW_SERVER packet. This packet includes the server's Server ID that it has created by itself and other relevant information about the server. The router receiving the ID MUST verify that the IP address in the Server ID is same as the server's real IP address. After router has received the SILC_PACKET_NEW_SERVER packet it distributes the information about newly registered server to all routers in the SILC network. More information about this is in [SILC2]. As the client needed to resolve the destination ID this MUST be done by the server that connected to the router, as well. The way to resolve it is to get the ID from previously received packet. The server MAY also use SILC_COMMAND_INFO command to resolve the ID. Server MUST also start Riikonen [Page 38] Internet Draft 29 July 2003 using its own Server ID as Source ID in SILC Packet Header and the router's Server ID as Destination when communicating with the router. 4.2.1 Announcing Clients, Channels and Servers After server or router has connected to the remote router, and it already has connected clients and channels it MUST announce them to the router. If the server is router server, also all the local servers in the cell MUST be announced. All clients are announced by compiling a list of ID Payloads into the SILC_PACKET_NEW_ID packet. All channels are announced by compiling a list of Channel Payloads into the SILC_PACKET_NEW_CHANNEL packet. Channels' mode and founder public key and other channel mode specific data is announced by sending SILC_NOTIFY_TYPE_CMODE_CHANGE notify list. Also, the channel users on the channels must be announced by compiling a list of Notify Payloads with the SILC_NOTIFY_TYPE_JOIN notify type into the SILC_PACKET_NOTIFY packet. The users' modes on the channel must also be announced by compiling list of Notify Payloads with the SILC_NOTIFY_TYPE_CUMODE_CHANGE notify type into the SILC_PACKET_NOTIFY packet. The router MUST also announce the local servers by compiling list of ID Payloads into the SILC_PACKET_NEW_ID packet. Also, clients' modes (user modes in SILC) MUST be announced. This is done by compiling a list of Notify Payloads with SILC_NOTIFY_UMODE_CHANGE notify type into the SILC_PACKET_NOTIFY packet. Also, channels' topics MUST be announced by compiling a list of Notify Payloads with the SILC_NOTIFY_TOPIC_SET notify type into the SILC_PACKET_NOTIFY packet. The router which receives these lists MUST process them and broadcast the packets to its primary router. When processing the announced channels and channel users the router MUST check whether a channel exists already with the same name. If channel exists with the same name it MUST check whether the Channel ID is different. If the Channel ID is different the router MUST send the notify type SILC_NOTIFY_TYPE_CHANNEL_CHANGE to the server to force the channel ID change to the ID the router has. If the mode of the channel is different the router MUST send the notify type SILC_NOTIFY_TYPE_CMODE_CHANGE to the server to force the mode change to the mode that the router has. The router MUST also generate new channel key and distribute it to the channel. The key MUST NOT be generated if the SILC_CMODE_PRIVKEY mode is set. If the channel has channel founder already on the router, the router Riikonen [Page 39] Internet Draft 29 July 2003 MUST send the notify type SILC_NOTIFY_TYPE_CUMODE_CHANGE to the server to force the mode change for the channel founder on the server. The channel founder privileges MUST be removed. The router processing the channels MUST also compile a list of Notify Payloads with the SILC_NOTIFY_TYPE_JOIN notify type into the SILC_PACKET_NOTIFY and send the packet to the server. This way the server (or router) will receive the clients on the channel that the router has. 4.3 Joining to a Channel This section describes the procedure when client joins to a channel. Client joins to channel by sending command SILC_COMMAND_JOIN to the server. If the receiver receiving join command is normal server the server MUST check its local list whether this channel already exists locally. This would indicate that some client connected to the server has already joined to the channel. If this is the case, the client is joined to the channel, new channel key is created and information about newly joined channel is sent to the router. The router is informed by sending SILC_NOTIFY_TYPE_JOIN notify type. The notify type MUST also be sent to the local clients on the channel. The new channel key is also sent to the router and to local clients on the channel. If the channel does not exist in the local list the client's command MUST be sent to the router which will then perform the actual joining procedure. When server receives the reply to the command from the router it MUST be sent to the client which sent the command originally. Server will also receive the channel key from the server that it MUST send to the client which originally requested the join command. The server MUST also save the channel key. If the receiver of the join command is router it MUST first check its local list whether anyone in the cell has already joined to the channel. If this is the case, the client is joined to the channel and reply is sent to the client. If the command was sent by server the command reply is sent to the server which sent it. Then the router MUST also create new channel key and distribute it to all clients on the channel and all servers that have clients on the channel. Router MUST also send the SILC_NOTIFY_TYPE_JOIN notify type to local clients on the channel and to local servers that have clients on the channel. If the channel does not exist on the router's local list it MUST check the global list whether the channel exists at all. If it does the client is joined to the channel as described previously. If the channel does not exist the channel is created and the client is joined to the channel. The channel key is also created and Riikonen [Page 40] Internet Draft 29 July 2003 distributed as previously described. The client joining to the created channel is made automatically channel founder and both channel founder and channel operator privileges are set for the client. If the router created the channel in the process, information about the new channel MUST be broadcast to all routers. This is done by broadcasting SILC_PACKET_NEW_CHANNEL packet to the router's primary route. When the router joins the client to the channel it MUST also send information about newly joined client to all routers in the SILC network. This is done by broadcasting the SILC_NOTIFY_TYPE_JOIN notify type to the router's primary route. It is important to note that new channel key is created always when new client joins to channel, whether the channel has existed previously or not. This way the new client on the channel is not able to decrypt any of the old traffic on the channel. Client which receives the reply to the join command MUST start using the received Channel ID in the channel message communication thereafter. Client also receives the key for the channel in the command reply. Note that the channel key is never generated or distributed if the SILC_CMODE_PRIVKEY mode is set. 4.4 Channel Key Generation Channel keys are created by router which creates the channel by taking enough randomness from cryptographically strong random number generator. The key is generated always when channel is created, when new client joins a channel and after the key has expired. Key could expire for example in an hour. The key MUST also be re-generated whenever some client leaves a channel. In this case the key is created from scratch by taking enough randomness from the random number generator. After that the key is distributed to all clients on the channel. However, channel keys are cell specific thus the key is created only on the cell where the client, which left the channel, exists. While the server or router is creating the new channel key, no other client may join to the channel. Messages that are sent while creating the new key are still processed with the old key. After server has sent the SILC_PACKET_CHANNEL_KEY packet client MUST start using the new key. If server creates the new key the server MUST also send the new key to its router. See [SILC2] for more information about how channel messages must be encrypted and decrypted when router is processing them. If the key changes very often due to joining traffic on the channel it is RECOMMENDED that client implementation would cache some of the old channel keys for short period of time so that it is able to decrypt all channel messages it receives. It is possible that on a heavy traffic Riikonen [Page 41] Internet Draft 29 July 2003 channel a message encrypted with channel key that was just changed is received by client after the new key was set into use. This is possible because not all clients may receive the new key at the same time, and may still be sending messages encrypted with the old key. When client receives the SILC_PACKET_CHANNEL_KEY packet with the Channel Key Payload it MUST process the key data to create encryption and decryption key, and to create the HMAC key that is used to compute the MACs of the channel messages. The processing is as follows: channel_key = raw key data HMAC key = hash(raw key data) The raw key data is the key data received in the Channel Key Payload. The hash() function is the hash function used in the HMAC of the channel. Note that the server also MUST save the channel key. 4.5 Private Message Sending and Reception Private messages are sent point to point. Client explicitly destine a private message to specific client that is delivered to only to that client. No other client may receive the private message. The receiver of the private message is destined in the SILC Packet Header as in any other packet as well. The Source ID in the SILC Packet Header MUST be the ID of the sender of the message. If the sender of a private message does not know the receiver's Client ID, it MUST resolve it from server. There are two ways to resolve the client ID from server; it is RECOMMENDED that client implementations send SILC_COMMAND_IDENTIFY command to receive the Client ID. Client MAY also send SILC_COMMAND_WHOIS command to receive the Client ID. If the sender has received earlier a private message from the receiver it should have cached the Client ID from the SILC Packet Header. If server receives a private message packet which includes invalid destination Client ID the server MUST send SILC_NOTIFY_TYPE_ERROR notify to the client with error status indicating that such Client ID does not exist. See [SILC2] for description of private message encryption and decryption process. 4.6 Private Message Key Generation Private message MAY be protected with a key generated by the client. The key may be generated and sent to the other client by sending packet Riikonen [Page 42] Internet Draft 29 July 2003 SILC_PACKET_PRIVATE_MESSAGE_KEY which travels through the network and is secured by session keys. After that the private message key is used in the private message communication between those clients. The key sent inside the payload SHOULD be randomly generated. This packet MUST NOT be used to send pre-shared keys. Another choice is to entirely use keys that are not sent through the SILC network at all. This significantly adds security. This key could be a pre-shared key that is known by both of the clients. Both agree about using the key and start sending packets that indicate that the private message is secured using private message key. In case of pre-shared keys (static keys) the IV used in encryption SHOULD be chosen randomly. It is also possible to negotiate fresh key material by performing Key Agreement. The SILC_PACKET_KEY_AGREEMENT packet MAY be used to negotiate the fresh key material. In this case the resulting key material is used to secure the private messages. Also, the IV used in encryption is used as defined in [SILC3], unless otherwise stated by the encryption mode used. By performing Key Agreement the clients may negotiate the cipher and HMAC to be used in the private message encryption and to negotiate additional security parameters. If the key is pre-shared key or other key material not generated by Key Agreement, then the key material SHOULD be processed as defined in [SILC3]. The hash function to be used SHOULD be SHA1. In the processing, however, the HASH, as defined in [SILC3] MUST be ignored. After processing the key material it is employed as defined in [SILC3]. In this case also, implementations SHOULD use the SILC protocol's mandatory cipher and HMAC in private message encryption. 4.7 Channel Message Sending and Reception Channel messages are delivered to a group of users. The group forms a channel and all clients on the channel receives messages sent to the channel. The Source ID in the SILC Packet Header MUST be the ID of the sender of the message. Channel messages are destined to a channel by specifying the Channel ID as Destination ID in the SILC Packet Header. The server MUST then distribute the message to all clients on the channel by sending the channel message destined explicitly to a client on the channel. However, the Destination ID MUST still remain as the Channel ID. If server receives a channel message packet which includes invalid destination Channel ID the server MUST send SILC_NOTIFY_TYPE_ERROR notify to the sender with error status indicating that such Channel ID Riikonen [Page 43] Internet Draft 29 July 2003 does not exist. See the [SILC2] for description of channel message routing for router servers, and channel message encryption and decryption process. 4.8 Session Key Regeneration Session keys MUST be regenerated periodically, say, once in an hour. The re-key process is started by sending SILC_PACKET_REKEY packet to other end, to indicate that re-key must be performed. The initiator of the connection SHOULD initiate the re-key. If perfect forward secrecy (PFS) flag was selected in the SILC Key Exchange protocol [SILC3] the re-key MUST cause new key exchange with SKE protocol. In this case the protocol is secured with the old key and the protocol results to new key material. See [SILC3] for more information. After the SILC_PACKET_REKEY packet is sent the sender will perform the SKE protocol. If PFS flag was set the resulted key material is processed as described in the section Processing the Key Material in [SILC3]. The difference with re-key in the processing is that the initial data for the hash function is just the resulted key material and not the HASH as it is not computed at all with re-key. Other than that, the key processing it equivalent to normal SKE negotiation. If PFS flag was not set, which is the default case, then re-key is done without executing SKE protocol. In this case, the new key is created by providing the current sending encryption key to the SKE protocol's key processing function. The process is described in the section Processing the Key Material in [SILC3]. The difference in the processing is that the initial data for the hash function is the current sending encryption key and not the SKE's KEY and HASH values. Other than that, the key processing is equivalent to normal SKE negotiation. After both parties have regenerated the session key, both MUST send SILC_PACKET_REKEY_DONE packet to each other. These packets are still secured with the old key. After these packets, the subsequent packets MUST be protected with the new key. 4.9 Command Sending and Reception Client usually sends the commands in the SILC network. In this case the client simply sends the command packet to server and the server processes it and replies with command reply packet. See the [SILC4] for detailed description of all commands. Riikonen [Page 44] Internet Draft 29 July 2003 However, if the server is not able to process the command, it is sent to the server's router. This is case for example with commands such as SILC_COMMAND_JOIN and SILC_COMMAND_WHOIS commands. However, there are other commands as well [SILC4]. For example, if client sends the WHOIS command requesting specific information about some client the server must send the WHOIS command to router so that all clients in SILC network are searched. The router, on the other hand, sends the WHOIS command further to receive the exact information about the requested client. The WHOIS command travels all the way to the server which owns the client and it replies with command reply packet. Finally, the server which sent the command receives the command reply and it must be able to determine which client sent the original command. The server then sends command reply to the client. Implementations should have some kind of cache to handle, for example, WHOIS information. Servers and routers along the route could all cache the information for faster referencing in the future. The commands sent by server may be sent hop by hop until someone is able to process the command. However, it is preferred to destine the command as precisely as it is possible. In this case, other routers en route MUST route the command packet by checking the true sender and true destination of the packet. However, servers and routers MUST NOT route command reply packets to clients coming from other servers. Client MUST NOT accept command reply packet originated from anyone else but from its own server. 4.10 Closing Connection When remote client connection is closed the server MUST send the notify type SILC_NOTIFY_TYPE_SIGNOFF to its primary router and to all channels the client was joined. The server MUST also save the client's information for a period of time for history purposes. When remote server or router connection is closed the server or router MUST also remove all the clients that was behind the server or router from the SILC Network. The server or router MUST also send the notify type SILC_NOTIFY_TYPE_SERVER_SIGNOFF to its primary router and to all local clients that are joined on the same channels with the remote server's or router's clients. Router server MUST also check whether some client in the local cell is watching for the nickname this client has, and send the SILC_NOTIFY_TYPE_WATCH to the watcher, unless the client which left the network has the SILC_UMODE_REJECT_WATCHING user mode set. Riikonen [Page 45] Internet Draft 29 July 2003 4.11 Detaching and Resuming a Session SILC protocol provides a possibility for a client to detach itself from the network without actually signing off from the network. The client connection to the server is closed but the client remains as valid client in the network. The client may then later resume its session back from any server in the network. When client wishes to detach from the network it MUST send the SILC_COMMAND_DETACH command to its server. The server then MUST set SILC_UMODE_DETACHED mode to the client and send SILC_NOTIFY_UMODE_CHANGE notify to its primary router, which then MUST broadcast it further to other routers in the network. This user mode indicates that the client is detached from the network. Implementations MUST NOT use the SILC_UMODE_DETACHED flag to determine whether a packet can be sent to the client. All packets MUST still be sent to the client even if client is detached from the network. Only the server that originally had the active client connection is able to make the decision after it notices that the network connection is not active. In this case the default case is to discard the packet. The SILC_UMODE_DETACHED flag cannot be set by client itself directly with SILC_COMMAND_UMODE command, but only implicitly by sending the SILC_COMMAND_DETACH command. The flag also cannot be unset by the client, server or router with SILC_COMMAND_UMODE command, but only implicitly by sending and receiving the SILC_PACKET_RESUME_CLIENT packet. When the client wishes to resume its session in the SILC Network it connects to a server in the network, which MAY also be a different from the original server, and performs normal procedures regarding creating a connection as described in section 4.1. After the SKE and the Connection Authentication protocols has been successfully completed the client MUST NOT send SILC_PACKET_NEW_CLIENT packet, but MUST send SILC_PACKET_RESUME_CLIENT packet. This packet is used to perform the resuming procedure. The packet MUST include the detached client's Client ID, which the client must know. It also includes Authentication Payload which includes signature computed with the client's private key. The signature is computed as defined in the section 3.9.1. Thus, the authentication method MUST be based in public key authentication. When server receive the SILC_PACKET_RESUME_CLIENT packet it MUST do the following: Server checks that the Client ID is valid client and that it has the SILC_UMODE_DETACHED mode set. Then it verifies the Authentication Payload with the detached client's public key. If it does not have the public key it retrieves it by sending SILC_COMMAND_GETKEY command to the server that has the public key from Riikonen [Page 46] Internet Draft 29 July 2003 the original client connection. The server MUST NOT use the public key received in the SKE protocol for this connection. If the signature is valid the server unsets the SILC_UMODE_DETACHED flag, and sends the SILC_PACKET_RESUME_CLIENT packet to its primary router. The routers MUST broadcast the packet and unset the SILC_UMODE_DETACHED flag when the packet is received. If the server is router server it also MUST send the SILC_PACKET_RESUME_CLIENT packet to the original server whom owned the detached client. The servers and routers that receives the SILC_PACKET_RESUME_CLIENT packet MUST know whether the packet already has been received for the client. It is a protocol error to attempt to resume the client session from more than one server. The implementations could set internal flag that indicates that the client is resumed. If router receive SILC_PACKET_RESUME_CLIENT packet for client that is already resumed the client MUST be killed from the network. This would indicate that the client is attempting to resume the session more than once which is a protocol error. In this case the router sends SILC_NOTIFY_TYPE_KILLED to the client. All routers that detect the same situation MUST also send the notify for the client. The servers and routers that receive the SILC_PACKET_RESUME_CLIENT must also understand that the client may not be found behind the same server that it originally came from. They must update their caches according to this. The server that now owns the client session MUST check whether the Client ID of the resumed client is based on the server's Server ID. If it is not it creates a new Client ID and send SILC_NOTIFY_TYPE_NICK_CHANGE to the network. It MUST also send the channel keys of all channels that the client has joined to the client since it does not have them. Whether the Client ID was changed or not the server MUST send SILC_PACKET_NEW_ID packet to the client. Only after this is the client resumed back to the network and may start sending packets and messages. It is also possible that the server did not know about the global channels before the client resumed. In this case it joins the client to the channels, generates new channel keys and distributes the keys to the channels as described in section 4.4. It is an implementation issue for how long servers keep detached client sessions. It is RECOMMENDED that the detached sessions would be persistent as long as the server is running. 5 Security Considerations Security is central to the design of this protocol, and these security considerations permeate the specification. Common security considerations Riikonen [Page 47] Internet Draft 29 July 2003 such as keeping private keys truly private and using adequate lengths for symmetric and asymmetric keys must be followed in order to maintain the security of this protocol. Special attention must also be paid to the servers and routers that are running the SILC service. The SILC protocol's security depends greatly on the security and the integrity of the servers and administrators that are running the service. It is recommended that some form of registration is required by the server and router administrator prior to acceptance to the SILC Network. Even though the SILC protocol is secure in a network of mutual distrust between clients, servers, routers and administrators of the servers, the client should be able to trust the servers they are using if they wish to do so. It however must be noted that if the client requires absolute security by not trusting any of the servers or routers in the SILC Network, it can be accomplished by negotiating private keys outside the SILC Network, either using SKE or some other key exchange protocol, or to use some other external means for distributing the keys. This applies for all messages, private messages and channel messages. It is important to note that SILC, like any other security protocol, is not a foolproof system; the SILC servers and routers could very well be compromised. However, to provide an acceptable level of security and usability for end users, the protocol uses many times session keys or other keys generated by the servers to secure the messages. This is an intentional design feature to allow ease of use for end users. This way the network is still usable, and remains encrypted even if the external means of distributing the keys is not working. The implementation, however, may like to not follow this design feature, and may always negotiate the keys outside SILC network. This is an acceptable solution and many times recommended. The implementation still must be able to work with the server generated keys. If this is unacceptable for the client or end user, the private keys negotiated outside the SILC Network should always be used. In the end it is the implementor's choice whether to negotiate private keys by default or whether to use the keys generated by the servers. It is also recommended that router operators in the SILC Network would form a joint forum to discuss the router and SILC Network management issues. Also, router operators along with the cell's server operators should have a forum to discuss the cell management issues. 6 References [SILC2] Riikonen, P., "SILC Packet Protocol", Internet Draft, Riikonen [Page 48] Internet Draft 29 July 2003 May 2002. [SILC3] Riikonen, P., "SILC Key Exchange and Authentication Protocols", Internet Draft, May 2002. [SILC4] Riikonen, P., "SILC Commands", Internet Draft, May 2002. [IRC] Oikarinen, J., and Reed D., "Internet Relay Chat Protocol", RFC 1459, May 1993. [IRC-ARCH] Kalt, C., "Internet Relay Chat: Architecture", RFC 2810, April 2000. [IRC-CHAN] Kalt, C., "Internet Relay Chat: Channel Management", RFC 2811, April 2000. [IRC-CLIENT] Kalt, C., "Internet Relay Chat: Client Protocol", RFC 2812, April 2000. [IRC-SERVER] Kalt, C., "Internet Relay Chat: Server Protocol", RFC 2813, April 2000. [SSH-TRANS] Ylonen, T., et al, "SSH Transport Layer Protocol", Internet Draft. [PGP] Callas, J., et al, "OpenPGP Message Format", RFC 2440, November 1998. [SPKI] Ellison C., et al, "SPKI Certificate Theory", RFC 2693, September 1999. [PKIX-Part1] Housley, R., et al, "Internet X.509 Public Key Infrastructure, Certificate and CRL Profile", RFC 2459, January 1999. [Schneier] Schneier, B., "Applied Cryptography Second Edition", John Wiley & Sons, New York, NY, 1996. [Menezes] Menezes, A., et al, "Handbook of Applied Cryptography", CRC Press 1997. [OAKLEY] Orman, H., "The OAKLEY Key Determination Protocol", RFC 2412, November 1998. [ISAKMP] Maughan D., et al, "Internet Security Association and Key Management Protocol (ISAKMP)", RFC 2408, November 1998. Riikonen [Page 49] Internet Draft 29 July 2003 [IKE] Harkins D., and Carrel D., "The Internet Key Exchange (IKE)", RFC 2409, November 1998. [HMAC] Krawczyk, H., "HMAC: Keyed-Hashing for Message Authentication", RFC 2104, February 1997. [PKCS1] Kalinski, B., and Staddon, J., "PKCS #1 RSA Cryptography Specifications, Version 2.0", RFC 2437, October 1998. [RFC2119] Bradner, S., "Key Words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [RFC2279] Yergeau, F., "UTF-8, a transformation format of ISO 10646", RFC 2279, January 1998. [RFC1321] Rivest R., "The MD5 Message-Digest Algorithm", RFC 1321, April 1992. [RFC3174] Eastlake, F., et al., "US Secure Hash Algorithm 1 (SHA1)", RFC 3174, September 2001. [PKCS7] Kalinski, B., "PKCS #7: Cryptographic Message Syntax, Version 1.5", RFC 2315, March 1998. [RFC2253] Wahl, M., et al., "Lightweight Directory Access Protocol (v3): UTF-8 String Representation of Distinguished Names", RFC 2253, December 1997. 7 Author's Address Pekka Riikonen Snellmaninkatu 34 A 15 70100 Kuopio Finland EMail: priikone@iki.fi 8 Full Copyright Statement Copyright (C) The Internet Society (2003). All Rights Reserved. This document and translations of it may be copied and furnished to others, and derivative works that comment on or otherwise explain it or assist in its implementation may be prepared, copied, published and distributed, in whole or in part, without restriction of any kind, provided that the above copyright notice and this paragraph are Riikonen [Page 50] Internet Draft 29 July 2003 included on all such copies and derivative works. However, this document itself may not be modified in any way, such as by removing the copyright notice or references to the Internet Society or other Internet organizations, except as needed for the purpose of developing Internet standards in which case the procedures for copyrights defined in the Internet Standards process must be followed, or as required to translate it into languages other than English. The limited permissions granted above are perpetual and will not be revoked by the Internet Society or its successors or assigns. This document and the information contained herein is provided on an "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Riikonen [Page 51]