ML-Based Detector for Microarchitectural and Side-Channel Attacks on Embedded IoT Devices

The use of IoT and embedded systems has been rapidly expanding in smart homes, healthcare, and industry, However, these technologies minimal computational and memory resources, lightweight design, and constant connectivity make them extremely vulnerable to side-channel and microarchitectural attacks. Existing countermeasures such as traditional security-based mechanisms can be too resourceheavy, thus reducing efficiency, performance, and increasing power consumption. In this proposed project, we will develop lightweight Machine Learning (ML)-based detectors that can run from the device on low-resource devices to identify and notify attack patterns in real time. The Detector uses simple featuressuch as execution timing, performance counters, and (optional) power bursts analysis using small ML models- Decision trees and lightweight NNs. This framework of detection will provide effective accuracy with very limited memory usage and energy overhead. The neutral design framework and methodology will offer tangible large-scale security expansion to provide practical, cost effective, defendable attack surface detection from cache contention, timing leakage, and covert channel attack detection.