Machine Learning-Based Autonomous Physical Security Defences
Subba Rao Polamurı, Knvpsb Ramesh, Kodathala Srihitha, M. Srıdevı, M. Sangeetha, A. Yv. M. Gurudatta · Advances in computer science research · 2024
Nearly 50 billion linked devices by 2025 will make physical entry to the target system much easier for attackers.The proliferation of embedded devices in mission-critical infrastructure and industrial control systems, as well as the existence of the Internet of Battlefield Things (IoBT), heighten this risk.Existing anti-tamper designs have limited efficacy in preventing specific types of attacks and rely on predetermined responses to detect manipulation, which can undermine system reliability.More covert attacks are now feasible thanks to new physical inspection technology.Therefore, there is an immediate need for improved defences that can endure the anticipated rise in hostile capabilities for a considerable amount of time.If we want to take physical security to the next level, this study suggests building a smart anti-tamper with machine learning algorithms.It employs a number of analytical frameworks, one of which can distinguish between normal functioning, known attack vectors, and unusual behaviour.To further aid in the reduction of false alarms and enhancement of operating time, the system has a tiered reaction mechanism as well as a recovery strategy.