11.3. Node, Attack, and Monitor Models
11.3.1. Node Model
A node in the network performs routing and switching. With the switching capability, a node can propagate a crosstalk attack from one attack channel to other normal connections. Some nodes can also support monitoring capabilities. Such a node is referred to as a monitor node. Otherwise, it is a nonmonitor node.
11.3.2. Crosstalk Attack Model
As shown in Figure 11.1, a crosstalk attack connection usually only affects another connection on the same wavelength. We define the following terms to describe a crosstalk attack model:
Upstream and downstream neighbor nodes. For a node on a certain path, its upstream neighbor node (UNN) is the previous node on that path. Similarly, its downstream neighbor node (DNN) is the next node on that path. UNN (node A, connection C) denotes the UNN of node A on connection C. Similarly, DNN (node A, connection C) denotes the DNN of node A on connection C.
A connection causing crosstalk attack results in different types of nodes and flows as follows:
The original attack flow (OAF) has a much higher energy level than permitted on a normal connection. The leakage of energy at a switch from the attack connection influences all other normal connections using the same wavelength on other fibers. The ability of an OAF to influence normal connections is the same at every node on its path. A node is called a primary attacked node (PAN) if there is an OAF originating at, terminating at, or passing through this node.
A normal connection sharing a node with an OAF is called a secondary attacked flow (SAF). The SAF has limited attack capability. If a normal connection C gets affected by an OAF at node u, then the connection C has attack capability only at node DNN (u, C), and we call DNN (u, C) a secondary attacked node (SAN).
A normal connection influenced by an SAF is called a final attacked flow (FAF). The FAF does not have the attack propagation capability.
A connection not affected by either OAF or SAF is called an attack-free flow (AFF). Similarly, a node that is neither a PAN nor a SAN is called an attack-free node (AFN). The union of AFF, SAF, and FAF is called an innocent flow (IF) set.
We illustrate the conncetion types by an example. As shown in Figure 11.3, connection C1 is the OAF, connection C2 is an SAF, connection C3 is an FAF, and connection C4 is an AFF. Nodes 1, 2, 3, and 6 are PANs. Node 5 is a SAN. The rest—nodes 4, 7, 8, and 9—are AFNs. Connection C1 can propagate its attack to connection C3 by affecting connection C2. According to this, it is expected that the OAF pollutes any connections passing through the PAN, and the SAF pollutes any normal connections passing through a SAN. Connections C2, C3, and C4 comprise the IF set.
Since the OAF, SAF, and FAF have different attack capabilities, the power level of these connection channels are as follows:
P(OAF)>> P(SAF)>P(FAF)>P(AFF)
where P(OAF) denotes the power level of OAF, and so on. For example, as shown in Figure 11.3, P(C1) > P(C2) >P(C3)>P(C4).
Figure 11.3. Example of attack flow and affected flow.
11.3.3. Monitor Node Model
A monitor needs to be as simple and cheap as possible. Because crosstalk attacks only change the optical power of normal signals, we only need the crosstalk detection method to detect the change in signal power, however, more than that seems unnecessary. Detection of the power level method is used as the core technique for monitoring. The following describes the monitor model in detail.
A monitor node can monitor all traffic passing through it, including the traffic that originates/terminates at the node.
The monitor node can detect the input/output connection power in all parts, including its demultiplexer, multiplexer, and switch plane, and distinguish them as needed to identify OAF, SAF, and FAF. We also use power detection methods to monitor the input and output connection signal power levels on all wavelengths in the input and the output fibers, as shown in Figure 11.4.
Figure 11.4. Attack monitoring mechanism for selective wavelength switches.
A connection can be in an attack/nonattack status at a monitor. We use A/Ā to indicate the attack/nonattack status of the connection.
It is possible that a monitor node may have multiple attacked connections passing through it. We consider three possibilities and the corresponding three responses from a monitor respectively:
One connection is an OAF while all the others are SAFs. Because P(OAF) > P(SAF), the monitor considers only the OAF connection to have attack capability. Thus, we assume that only the state of the OAF is set as A, while the other SAFs states are set as Ā.
More than one connection is an SAF, but none is an OAF. In this situation, the monitor can detect several connections that have similar unexpected high power. We assume that the monitor sets all SAFs to A state.
Two or more connections are OAFs. In this situation, similar to step 1 above, the monitor can detect several connections that have high power, and set states of these connections to A and sets states of other connections to Ā.
Figure 11.5 shows a 3×3 mesh network. Connections C1 and C6 are two OAFs. Nodes 2, 4, 6, and 8 are monitors. On node 2, because connection C1 and C6 are two OAFs passing through this monitor, connections C1 and C6’s statuses are set as A, while connection C3 is set as Ā because it is an SAF. On node 4, both connections C2 and C4 are SAFs, and no OAF passes through node 4. Thus, node 4 sets both C2 and C4 as A. On node 6, OAF connection C1 is set as A while C3 and C4 are set as Ā. On node 8, because C2 does not have attack capability on this node, both C2 and C5 are set as Ā.
Figure 11.5. Different attack connections passing through monitors.
