Adversarial Challenges in Distributed AI

Abstract

With a special emphasis on distributed AI/ML systems, the abstract explores the intricate world of adversarial challenges in 6G networks. With their unmatched capabilities—like ultra-high data rates and ultra-low latency it highlights the crucial role that 6G networks will play in determining the direction of communication in the future. Still, it also highlights the weaknesses in these networks' distributed AI/ML systems, emphasizing the need for strong security measures to ward off potential attacks. Also covered in the abstract is the variety of adversarial threats that exist, such as model evasion, data poisoning, backdoors, membership inference, and model inversion attacks.