Journal of Network and Computer Applications
Peer-to-peer data mining classifiers for decentralized detection of network attacks
ADC '13 Proceedings of the Twenty-Fourth Australasian Database Conference - Volume 137
| Hi-index | 0.00 |
The work presents a decentralized protocol that allows self-organization of autonomous wireless devices in mobile mesh networks. In the resulting infrastructure the routing and information services are provided in a peer-to- peer fashion. Both services are performed through multi- hop radio transmissions among participant nodes with no fixed infrastructure required and avoiding to use broadcast of messages. Since links among nodes may be continuously destroyed and created, as a consequence of nodes movement and/or variations in connectivity range (e.g. due to temporary obstacles), our protocol is based on virtual network addresses which are dynamically assigned to groups of nodes (called islands). Routing requests are performed through unicast transmissions when traveling among islands and exploiting wireless overhearing when involving members of the same island (in- island routing). We will demonstrate that assigning virtual coordinates to groups of nodes guarantees network flexibility with a reasonable amount of overheard and that the amount of radio transmissions required is scalable with the number of nodes. Simulations have shown that the proposed solution can generate reliable networks, despite the unpredictable topology, assuming the use of devices with common radio range connectivity moving at pedestrian velocity.