A Novel Framework for Smart Cyber defence: A Deep-Dive into Deep Learning Attacks and defences

Deep learning techniques have been widely adopted for cyber defence applications such as malware detection and anomaly detection. The ever-changing nature of threats has made a constantly evolving field. Smart manufacturing is critical to the broader thrust towards Industry 4.0 5.0. Developing advanced technologies in smart requires enabling paradigm shift manufacturing, while cyber-attacks significantly threaten manufacturing. For example, attack (e.g., backdoor) occurs during model’s training process. Cyber affects models impacts resultant output be misled. Therefore, this paper proposes novel comprehensive framework deep security. collectively incorporates threat model, data, model proposed encompasses multiple layers, including privacy protection data models. In addition statistical intelligent maintaining confidentiality, covers structural perspective, i.e., policies procedures securing data. study then offers different methods make robust against attacks coupled with model. Along security, helps defend systems by identifying potential or actual vulnerabilities putting countermeasures control place. Moreover, based on our analysis, provides taxonomy backdoor defences. addition, qualitative comparison existing Finally, highlights future directions defences possible way further research.