Multi-Scale Temporal Spike-Based Intrusion Detection: A Novel Framework for Cybersecurity Applications

Modern cyber threats exhibit complex temporal characteristics spanning microseconds to months, challenging traditional intrusion detection systems that operate within fixed time windows. This paper presents the Multi-Scale Temporal Spike-Based Intrusion Detection System (MSTS-IDS), leveraging Spiking Neural Networks (SNNs) for energy-efficient multi- temporal pattern recognition. Our hierarchical architecture pro- cesses network traffic across three temporal scales simultaneously: fast (s-ms) for port scans and DDoS attacks, medium (s- min) for behavioral patterns, and long (hr-months) for Advanced Persistent Threats (APTs). Experimental validation demonstrates 96.3 % detection accuracy with 75% energy reduction compared to traditional approaches. The framework achieves 93.7% APT campaign detection accuracy while maintaining sub-millisecond response times for fast attacks.