Adversarial AI in Cyber Security
Ajay B. Gadicha, Vijay B. Gadicha, Mohammad Moin Maniyar · Advances in information security, privacy, and ethics book series · 2025
Adversarial AI is rapidly emerging as one of the most pressing concerns in cybersecurity, posing significant threats to the reliability and security of machine learning models used across various sectors. This book provides an in-depth exploration of how adversaries manipulate AI systems to bypass security protocols, mislead machine learning algorithms, and exploit vulnerabilities in both the training and deployment phases of AI models. By introducing subtle and carefully designed perturbations to input data, adversarial attackers can cause misclassification, model inversion, and even gain unauthorized access to sensitive information. The widespread adoption of AI technologies in fields such as finance, healthcare, autonomous vehicles, and national security systems amplifies the risks associated with these attacks. This chapter delves into the various types of adversarial attacks including evasion attacks, poisoning attacks, and backdoor insertion demonstrating how they compromise system integrity.