Search papers, labs, and topics across Lattice.
The paper introduces Spike-PTSD, a novel adversarial attack on Spiking Neural Networks (SNNs) inspired by neural firing patterns observed in Post-Traumatic Stress Disorder (PTSD). Spike-PTSD identifies decision-critical layers and neurons based on hyper/hypoactivation, then crafts adversarial examples by optimizing spike scaling to mimic PTSD-like neural abnormalities. Experiments across diverse datasets, encoding schemes, and SNN architectures demonstrate that Spike-PTSD achieves over 99% attack success rates, highlighting a significant vulnerability in SNN robustness.
SNNs, often touted for their bio-plausibility and security, are alarmingly susceptible to adversarial attacks that mimic neural firing patterns in PTSD, achieving near-perfect attack success.
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.
