Learning Based Adaptive Network Immune Mechanism to Defense Eavesdropping Attacks

Encryption mechanisms improve the security of transmitting data. Nevertheless, attackers might silently eavesdrop packets to crack sensitive information or launch cyberattacks to damage the network performance of victims. In this paper, we propose a learning based adaptive network immune mechanism (LANIM) to prevent the eavesdropping attacks. Specifically, LANIM is equipped with three
defense lines
and one
constraint
. The first defense line focuses on making decisions about abnormal network conditions by the minimum risk machine learning algorithms. The second defense line is the encryption strategy which focuses on the intent and application. The programmable devices implement novel policies such as multipath transmission and packet encapsulation. LANIM inherits the existing countermeasures based on computational complexity, which is the third defense line. Besides, the policy dynamically updates with random seeds is the constraint of the cyberattack. The attackers have to conquer these three immune lines before the new policy update otherwise the offensive is shattered. We implement LANIM with the P4 language and the Smart Identifier Network (SINET) framework, evaluate the ability to resist the hazards of eavesdropping attacks and explore the trade-offs between security and performance.