Abstract:

Machine learning (ML), which has been studied since the 1950s, is a branch of artificial intelligence that allows machines to adapt and learn from data rather than being programmed. In more recent times, ML has been used in systems including robotics, autonomous vehicles, smart power grids, and process control. These kinds of systems directly affect human safety and life. Their ML models must therefore be protected from adversaries that aim to damage users or compromise their privacy. The many advantages that machine learning has provided for security and CPS/IoT, both generally and specifically, including the improvement of intrusion detection systems and decision accuracy in CPS/IoT. CPS stands for Cyber-Physical Systems. High-tech sensors combined with actual physical places make up cyber-physical systems (CPS). In the context of CPS systems, these intimate couplings of sensors with communication infrastructure that are inextricably linked to society's Crucial Infrastructures (C.I.) are more frequently observed. The use of adversarial machine learning research to cyber-physical systems (CPS) like autonomous vehicles and healthcare is examined in this paper. As a result, this study will provide as a springboard for further investigation into adversarial ML and CPSs. Providing a deeper grasp of this new trans disciplinary methodology is the goal of this study. The characteristics of CPSs are discussed.

Keywords: