Neuro‐Cognitive ML Framework for Deepfake Detection and Spoofed Sensor Integrity in Cyber–Physical Systems

Cyber–Physical Systems (CPS) are progressively dependent on artificial intelligence (AI)-powered sensing, perception and decision-making capabilities, thereby exposing them to sophisticated cyber-attacks that exploit the digital and physical layers. Two threats on the horizon – deepfake manipulation and spoofed sensor data – can severely compromise smart hospitals, autonomous vehicles, defense grids and other critical infrastructures. In this chapter, a neuro-cognitive machine learning framework is presented that combines the merits of deep neural networks with cognitive reasoning models to recognize, understand and alleviate the adversarial threats. Federated and distributed learning approaches that facilitate collaborative model training without sharing raw data are potential solutions to privacy and security issues in large-scale CPS deployments. The inclusion of explainable AI (XAI) in the cognitive layer will help in making decisions that are transparent and can be trusted in environments that are safety-critical.