Salem

We compare our solution with the solution presented in Salem et al. [37] because it is the solution that tackles the most similar topic to ours, which is that of selecting the most trustworthy AP. In the solution of Salem et al., each AP is characterised by a triplet (AQW, RQW, PW) where AQW is the QoS advertised by W where W represents the Wireless Internet Service Providers (WISPs);

RQW is the real QoS provided by W; and PW is the price that W is asking for. Salem et al. consider that a WISP’s (wireless Internet service provider) W is honest if it advertises the real QoS it is offering (i.e., RQW = AQW). W is misbehaving if it advertises a QoS that is higher that the real QoS that it is offering (i.e., RQW <

AQW). And it is modest if it advertises a QoS that is lower than the real QoS it is offering (i.e., RQW > AQW). Salem et al. initialise the reputation of the WISPs to maxRR = 100. At the end of each session, an MN (Mobile Node) sends to a TCA (Trusted Central Authority) its satisfaction level Sl = QoSEvalW where QoSEvalW = RQW/AQW. Each simulation lasts for 50,000 seconds and the reputation updates are made every 2,000 seconds. The new reputation RRW (t + 1) of W is computed as follows:

RRW (t + 1) = ß · RRW (t) + (1 − ß) ·feedbackW/nbSW where RRW (t) is the current reputation of W. nbSW is the number of sessions established by W (and already closed) during the last 2,000 seconds and the feedback is the sum of all QoSEvalW received over all these sessions (the absence of feedback is considered as QoSEvalW = 0 ). ß represents the “weight of the past” and is set to 1/2 in their simulations.

 APN have an RQW = AQW

 APM have an RQW < AQW.

 APS change between APN and APM.

Simulation 5: mix of UNs & APNs & APMs

Scenario 1: We have 230 users and 230 APs. In the 230 APs we have: 207 APNs and 23 APMs, 0 UMs and 230 UNs. All the APs at the beginning of each simulation have a trust value of 100.

Figure 15 Result of Simulation 5: mix of UNs & APNs & APMs

The number of normal users using APMs increases each time.

This is possible because there is not any threshold to prevent users from connecting to APMs. There are more users using APNs than users using APMs; so the solution of Salem et al. promotes the selection of APNs, but still the number of users using APMs is too high. By comparing it with our solution, we can notice that our solution decreases the number of users using APMs (Figure 13).

Simulation 6: mix of UNs & APMs

Scenario 2: We have 230 users: 230 APMs, 0 UMs and 230 UNs.

All the APs at the beginning of each simulation have a trust value of 100.

Figure 16 Result of Simulation 6: mix of UNs & APMs

Figure 16 shows how Salem’s solution behaves when there are only malicious APs. As we can see, with Salem’s solution, a lot of users use APMs because of the fact that Salem’s algorithm forces the use of APs who have the best reputation value among other APs; even if all the APs have a low reputation value. Our solution prevents this case from happening because the user will connect to an APM only if they are the first to use this AP after the insertion of the AP in the network (Figure 12). This is possible due to our threshold K, which prevents users from connecting to an AP which has a trust value lower than the threshold K.

Simulation 7: mix of UNs & APNs & APMs & APSs

Scenario 3: We have: 189 APNs, 19 APSs, 46 APMs, 0 UMs and 230 UNs. All the APs at the beginning of each simulation have a trust value of 100. In Salem’s solution, the APS will change its behaviour when its trust value is lower than 50.

Figure 17 Result of simulation 7: mix of UNs & APNs & APMs & APSs

Figure 17 shows how Salem’s solution deals with APSs and APMs at the same time. The result of simulation 7 shows that Salem’s solution does not prevent users from connecting to APMs and APSs. But we notice that just a few users used APSs and sometimes, the number of users using APSs was close to zero.

So, Salem’s solution is not useful for preventing users from using APSs and APMs.

Simulation 8: mix of UNs & APNs & APSs & UMs

Scenario 4: We have 211 APNs, 184 UNs, 46 UMs, 19 APSs and 0 APMs. All the APs at the beginning of each simulation have a trust value of 0.5 and all the users have a trust value of 0 with 10 friends. All the APs at the beginning of each simulation have a trust value of 100.

Figure 18 Result of Simulation 8: mix of UNs & APNs & APS & UMs

In this simulation, we introduce UMs and APSs at the same time.

As we can see in the first round, we have less than 10 malicious users using normal APs. This number remains more or less stable.

Salem’s solution does not deal with malicious users nor does it deal with APSs, as can be seen in the results of simulation 8. So Salem’s solution is not suitable for dealing with UMs and APSs.