Introduction to Computer Security - Section 10.3. Networks and Cryptography

Before we discuss Internet protocols, a review of the relevant properties of networks is in order. The ISO/OSI model [894] provides an abstract representation of networks suitable for our purposes. Recall that the ISO/OSI model is composed of a series of layers (see Figure 10-2). Each host, conceptually, has a principal at each layer that communicates with a peer on other hosts. These principals communicate with principals at the same layer on other hosts. Layer 1, 2, and 3 principals interact only with similar principals at neighboring (directly connected) hosts. Principals at layers 4, 5, 6, and 7 interact only with similar principals at the other end of the communication. (For convenience, "host" refers to the appropriate principal in the following discussion.)

Figure 10-2. The ISO/OSI model. The dashed arrows indicate peer-to-peer communication. For example, the transport layers are communicating with each other. The solid arrows indicate the actual flow of bits. For example, the transport layer invokes network layer routines on the local host, which invoke data link layer routines, which put the bits onto the network. The physical layer passes the bits to the next "hop," or host, on the path. When the message reaches the destination, it is passed up to the appropriate level.

Each host in the network is connected to some set of other hosts. They exchange messages with those hosts. If host nob wants to send a message to host windsor, nob determines which of its immediate neighbors is closest to windsor (using an appropriate routing protocol) and forwards the message to it. That host, baton, determines which of its neighbors is closest to windsor and forwards the message to it. This process continues until a host, sunapee, receives the message and determines that windsor is an immediate neighbor. The message is forwarded to windsor, its endpoint.

Definition 105. Let hosts C₀, …, C_n be such that C_i and C_i₊₁ are directly connected, for 0 i < n. A communications protocol that has C₀ and C_n as its endpoints is called an end-to-end protocol. A communications protocol that has C_j and C_j₊₁ as its endpoints is called a link protocol.

The difference between an end-to-end protocol and a link protocol is that the intermediate hosts play no part in an end-to-end protocol other than forwarding messages. On the other hand, a link protocol describes how each pair of intermediate hosts processes each message.

EXAMPLE: The telnet protocol is an applications layer protocol that allows users to obtain a virtual terminal on a remote host. Thus, it is an end-to-end protocol. IP is a network layer protocol that guides messages from a host to one of its immediate neighbors. Thus, it is a link protocol.

The protocols involved can be cryptographic protocols. If the cryptographic processing is done only at the source and at the destination, the protocol is an end-to-end protocol. If cryptographic processing occurs at each host along the path from source to destination, the protocol is a link protocol. When encryption is used with either protocol, we use the terms end-to-end encryption and link encryption, respectively.

EXAMPLE: If the messages between the telnet client and server are enciphered [915], the encipherment and decipherment occur at the client and the server only. The protocol uses end-to-end encryption. The PPP Encryption Control Protocol [627] enciphers messages between intermediate hosts. When a host gets the message, it deciphers the message, determines which neighbor to send it to, reenciphers the message using the key appropriate for that neighbor, and sends it to that neighbor. This protocol uses link encryption.

In link encryption, each host shares a cryptographic key with its neighbor. (If public key cryptography is used, each host has its neighbor's public key. Link encryption based on public keys is rare.) The keys may be set on a per-host basis or a per-host-pair basis. Consider a network with four hosts called windsor, stripe, facer, and seaview. Each host is directly connected to the other three. With keys distributed on a per-host basis, each host has its own key, making four keys in all. Each host has the keys for the other three neighbors, as well as its own. All hosts use the same key to communicate with windsor. With keys distributed on a per-host-pair basis, each host has one key per possible connection, making six keys in all. Unlike the per-host situation, in the per-host-pair case, each host uses a different key to communicate with windsor. The message is deciphered at each intermediate host, reenciphered for the next hop, and forwarded. Attackers monitoring the network medium will not be able to read the messages, but attackers at the intermediate hosts will be able to do so.

In end-to-end encryption, each host shares a cryptographic key with each destination. (Again, if the encryption is based on public key cryptography, each host hasor can obtainthe public key of each destination.) As with link encryption, the keys may be selected on a per-host or per-host-pair basis. The sending host enciphers the message and forwards it to the first intermediate host. The intermediate host forwards it to the next host, and the process continues until the message reaches its destination. The destination host then deciphers it. The message is enciphered throughout its journey. Neither attackers monitoring the network nor attackers on the intermediate hosts can read the message. However, attackers can read the routing information used to forward the message.

These differences affect a form of cryptanalysis known as traffic analysis. A cryptanalyst can sometimes deduce information not from the content of the message but from the sender and recipient. For example, during the Allied invasion of Normandy in World War II, the Germans deduced which vessels were the command ships by observing which ships were sending and receiving the most signals. The content of the signals was not relevant; their source and destination were. Similar deductions can reveal information in the electronic world.

EXAMPLE: ARS&C is an engineering firm developing the next generation of network protocols. Each employee of ARS&C has his or her own workstation. All network traffic is enciphered using end-to-end encryption. A competitor of the company appears to be obtaining proprietary data. ARS&C has hired Alice to figure out who is leaking the information.

Alice begins by monitoring all network traffic. She notices that the workstations are grouped into three different divisions: corporate management, sales, and engineering. The leaks are coming from the engineering systems. She looks at the sources and destinations of all connections to and from the engineering systems and notices that the connections from corporate management center on three systems: curly, larry, and moe. The connections from larry always occur between midnight and four in the morning; those from the other two occur during the day. Alice then looks at the events of the days on which the connections take place. The connections from curly and moe occur on the days of management reviews and are invariably to the ftp or www port. The connections from larry are more infrequent and are to the telnet port. A few days after each connection, the competitor seems to have acquired new proprietary information.

From this analysis, Alice suggests that the host larry is somehow involved in the problem. She needs to check the systems that larry connects to and see if the proprietary data is on those systems. At no time has Alice read any of the traffic, because it is encrypted; but from the traffic analysis, she has determined the system involved in the compromise.