A DDoS Attack Detection Method Based on Spiking Neural Networks
Tianhuai Yin, Aiqun Hu · 2025
With the annual advancement of computers, network threats have become increasingly severe. Among these, distributed denial of service (DDoS) attacks pose a significant threat. In recent years, DDoS attack detection methods have predominantly relied on machine learning approaches, which suffer from high computational and memory overhead and poor noise robustness. Additionally, traditional firewall-based defenses are inadequate to address the increasingly complex network environments. In recent years, the concept of bio-inspired security has emerged, drawing inspiration from the way biological neurons process external stimuli. This paper leverages this concept and proposes a DDoS attack detection method based on spiking neural networks (SNNs) within a bio-inspired security framework. Building upon the foundation of SNNs, the method uses a reused encoding scheme for spiking encoding and incorporates neural correlation to further reduce computational and memory overhead. To leverage existing labeled DDoS data, this paper introduces a semi-supervised model by adding the final layer on the SNNs. Experimental results demonstrate that the proposed SNNs-based DDoS attack detection method significantly reduces the computational and memory overhead required for training compared to several existing algorithms. Additionally, the noise robustness of the method has been notably enhanced.