A Collaborative Bayesian Watchdog for Detecting Black Holes in MANETs.

Watchdogs are a well-known mechanism to detect threats and attacks from misbehaved and selfish nodes in computer networks. In infrastructureless networks, such as MANETs, attack detection and reaction is a key issue to the whole network. Watchdog systems overhear traffic and perform analysis using data collected to decide the grade of misbehaviour that their neighbour nodes present, so accuracy and detection speed play a key role in achieving the right level of network security and performance. The problem behind the use of watchdogs is that they can cause a relatively high level of false positives and false negatives, so increasing their accuracy is also basic to successfully implement this technology in real MANET environments. This paper proposes a collaborative approach for detecting black holes and selfish nodes in MANETs, using a set of watchdogs which collaborate to enhance their individual and collective performance. The paper shows that using this collaborative watchdog approach the detection time of misbehaved nodes is reduced and the overall accuracy increased.