A trust-propagation-based scheme against pollution attacks in wireless network coding

Network coding is a new paradigm that provides possibility to maximize network throughput, decrease power consumption, and increase reliability. Forwarder nodes in network coding system receive input packets and mix them to generate output packets. Because of mixing nature of network coding packets, these systems are vulnerable to pollution attacks, in which attacker can inject polluted packets into the network or corrupt input packets to generate polluted output packets. Forged packets and corrupted packets generated by adversaries propagate quickly into network and infect a large proportion of network nodes in an epidemic way. Existing mechanisms addressing pollution attacks either have high computational overhead which decreases network coding performance or need an extra secure channel. In this paper, we propose a Trust-Propagation-based scheme to identify and filter pollution attacks and dynamically detect and remove adversaries. In our scheme, network is divided into regions. In each region, there is one trusted node which detects pollution attacks and drops corrupted packets. If a node sends a corrupted packet to a trusted node, this trusted node decreases its trust in the sender. The sender in turn decreases its trust in the upstream sender node and this process continues recursively. If the receiver node's trust in the sender one becomes smaller than is required, the sender node is known as an adversary. Experimental results show that our scheme can decrease computation overhead, latency and also increase throughput of the system significantly, and it's the first scheme that dynamically detects and removes adversaries.