This chapter explores the application of CNN accelerators in enhancing hardware security, with a specific focus on countermeasures for flooding fenial-of-service (FDoS) attacks in networks-on-chip (NoCs). It begins by introducing the role of deep learning in hardware security, highlighting key examples and applications. The chapter then presents a CNN-based framework for FDoS detection and localization, incorporating innovative techniques such as multi-frame fusion and victim completing enhancement for precise attack path, victim, and attacker localization. Feature selection, CNN model design, and hardware overhead considerations are discussed to balance detection accuracy and system efficiency. Experimental results demonstrate the framework’s effectiveness, achieving up to 95.8% detection accuracy and 91.7% localization accuracy on a 16 \(\times \) 16 NoC with minimal hardware overhead. Comparative analyses with related works showcase the framework’s scalability and adaptability, underscoring its potential for addressing NoC security challenges in large-scale systems.