ContagAlert: Using Contagion Theory for Adaptive, Distributed Alert Propagation

Large-scale distributed systems, e.g., grid or P2P networks, are targets for large-scale attacks. Unfortunately, few existing systems support propagation of alerts during the attack itself while also suppressing disruptive alerts from faulty or malicious sources. This paper proposes the "ContagAlert" protocol, which uses contagion spreading behavior to spread alerts. ContagAlert rapidly propagates alerts during attacks while also suppressing disruptive alerts. The core contagion protocols in the system are completely localized, but result in desired behavior at the network scale. We analyze and evaluate our protocol with synthetic simulations and in both Internet worm and DoS attack scenarios