Information Technology Reference
In-Depth Information
Fig. 2. DR and FPR vs. amount of clean data during the training
Now we will present the results of the experiments where the SOM algorithm is
being trained with the data that contains the traces of attacks. Bad nodes make 28.6%
of all the nodes. The results are presented in Fig.2. We can observe that the detection
rate decreases, as the amount of the clean data during the training also decreases,
while the false positive rate increases. This could be expected, as the “unclean” data
makes bigger part of the training data, it is harder to distinguish it.
5 Conclusions
In this work we have presented an approach for detecting bad mouthing attack on
reputation systems. The approach is based on outlier detection using SOM algorithm.
We have proven that it is capable of achieving 100% detection rate with 0% false
positive rate if trained with clean data and up to 28.6% of the nodes are malicious. For
the last case, when 28.6% of the nodes are malicious, the detection of the attack is
possible if at least 40% of the data are clean.
In the future we plan to test the approach on another colluding attack, ballot stuff-
ing, where a number of entities agree to give positive feedback on an entity (often
with adversarial intentions), resulting in it quickly gaining high reputation. In addi-
tion, we will test the performances of other clustering techniques instead of SOM,
such as Growing Neural Gas [13].
Acknowledgments. This work was funded by the Spanish Ministry of Industry, Tour-
ism and Trade, under Research Grant TSI-020301-2009-18 (eCID), the Spanish Min-
istry of Science and Innovation, under Research Grant TEC2009-14595-C02-01, and
the CENIT Project Segur@.
References
1. Moya, J.M., Araujo, A., Bankovic, Z., de Goyeneche, J.M., Vallejo, J.C., Malagon, P.,
Villanueva, D., Fraga, D., Romero, E., Blesa, J.: Improving security for SCADA sensor
networks with reputation systems and Self-Organizing maps. Sensors 9, 9380-9397 (2009)
2. Boukerch, A., Xu, L., EL-Khatib, K.: Trust-based Security for Wireless Ad Hoc and Sen-
sor Networks. Comput. Commun. 30, 2413-2427 (2007)
Search WWH ::




Custom Search