The later fact has the nice property to reduce the required signaling traffic of our proposed DTCF. Normally no A-Message about a fake event is transmitted from [Z.sub.1] unless the network is sparse and one and only one vehicle is in [Z.sub.d] that is malicious.
The excellent detection capability of DTCF is exacerbated when [P.sub.m] = 1, we see in Fig.
However the selfish behavior has no effect when malicious nodes use [P.sub.m] = 1, showing once again the DTCF excellent detection capability of malicious vehicles.
In contrast to the above, we now investigate the detection capability of our proposed DTCF per layer (within each layer) of the zone of interest [Z.sub.I].
We do not show the variance of [T.sub.m] = 0 and [T.sub.m] = 1 as they are null for different scenarios which clearly indicates that our proposed DTCF computes indeed exact trust metrics.
First of all, we observe for the case of [P.sub.m] = 1 the excellent full detection capability (100%) of our proposed DTCF in both cases of F = 1 and F=0.8.
Large delays are only obtained when malicious vehicles are allowed to choose very low values of [P.sub.m], otherwise our proposed DTCF insures a very short detection delay less even than 1 second.
17 that the consistency is virtually independent from both the number of considered events and the number of hops in [Z.sub.I] For [T.sub.m](v) equal 0 and 1, we get null values indicating the excellent capability of our proposed DTCF to compute exact and consistent trust metrics.
We compare the average trust metric computed by DTCF to that computed by the TBSE protocol proposed in .
18, that DTCF and TBSE compute the same value of trust metrics in the case of a continuous malicious behavior.
We here propose a qualitative comparison between our proposed DTCF and three other proposals ,  and .
Table 4 presents a qualitative comparison between DTCF and the three relevant approaches [6, 7, 13] using the above criteria along with the use of a recommendation system and the necessity of Road Side Units (RSU).