Adding Security to Networks-on-Chip using Neural Networks

Modern computing systems are using Networkson-Chip (NoCs) for scalable on-chip communications. Traditional security attacks have focused on the computing cores however modern attacks can focus on the NoC interconnect as a means to injecting unwanted traffic. This poses a significant security risk to physical systems that rely on sensory feedback and control signals. Malicious modification of these signals generates abnormal traffic patterns which affect the operation of the system and its performance. In this paper, we aim to identify abnormal traffic patterns (attacks) in Networks-on-Chip data through the use of Spiking Neural Networks. We explore the vulnerabilities of Denial-of-Service (DoS) attacks and report on evaluations to identify the impact of the duration of individual attacks on the rate of detection.