Spike-PTSD: A Bio-Plausible Adversarial Example Attack on Spiking Neural Networks via PTSD-Inspired Spike Scaling

2026-04-02

AI Summary

This work addresses the insufficient robustness of spiking neural networks (SNNs) against adversarial attacks and the neglect of their biologically plausible dynamic properties in existing methods. Inspired by the aberrant neuronal firing mechanisms observed in post-traumatic stress disorder (PTSD), this study introduces, for the first time, PTSD-related neural dynamics into SNN adversarial attacks. By identifying decision-critical layers, selecting neurons based on hyper- and hypo-activation features, and integrating spike-scaling strategies with a dual-objective optimization framework, the authors develop a biologically interpretable and general-purpose attack methodology compatible with diverse encoding schemes and SNN architectures. The proposed approach achieves over 99% attack success rates across six datasets, three encoding types, and four SNN models, systematically exposing the inherent vulnerabilities of SNNs.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness

Application Category

• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Spiking Neural Networks (SNNs) are energy-efficient and biologically plausible, ideal for embedded and security-critical systems, yet their adversarial robustness remains open. Existing adversarial attacks often overlook SNNs' bio-plausible dynamics. We propose Spike-PTSD, a biologically inspired adversarial attack framework modeled on abnormal neural firing in Post-Traumatic Stress Disorder (PTSD). It localizes decision-critical layers, selects neurons via hyper/hypoactivation signatures, and optimizes adversarial examples with dual objectives. Across six datasets, three encoding types, and four models, Spike-PTSD achieves over 99% success rates, systematically compromising SNN robustness. Code: https://github.com/bluefier/Spike-PTSD.

Problem

• Spiking Neural Networks
• adversarial robustness
• bio-plausible dynamics
• adversarial attacks
• PTSD-inspired

Innovation

• Spiking Neural Networks
• Adversarial Attack
• PTSD-inspired
• Bio-plausible
• Spike Scaling