Design and implementation of Denial-of-Service attack in network of multiprocessor systems-on-chip with anomaly detection approach

IoT architecture based on the multiprocessor system-on-chip (MPSoC) is widely used because of the third-party intellectual property cores (3PIPs) supplied globally. Therefore, 3PIPs present an intrinsic security risk. Decentralization of the network-on-chip (NoC) architecture and its diverse links lead to Network Vulnerability, with the Flooding denial-of-service (F-DoS) attack being the most common among many. This research aims to address the vulnerability of MPSoC designs to security threats, such as DoS attacks in NoC. The main aims include designing a real-time monitoring system based on the runtime machine learning, examining the system's accuracy in attack detection, and testing the drop in system performance. Including linear regression, decision trees, and support vector machines, these models are statically trained machine learning models that help realize objectives. Such models are designed to deal with the complex cases of NoC congestion patterns, which are highly different to various application mapping combinations. The study investigates crafty learning techniques to detect F-DoS attacks in NoC-based MPSoCs, providing the results depicted in the following as the effects. This study shows the high precision of machine learning models in detecting F-DoS attacks amid the constrained dynamic of stress NoC, where different attack scenarios were considered and analyzed. Moreover, as runtime effectiveness is almost unaffected, the feasibility of real-time intrusion detection is thus confirmed. This study shows that it can improve the safety of system-on-a-chip designs by offering a new technique for identifying and dealing with F-DoS attacks, resulting in absolute system performance resilience against incremental and evolving threats in the context of NoC-based MPSoCs.