PulseAugur
EN
LIVE 10:19:18

New attack targets event-based SNNs, increasing latency by 38%

By PulseAugur Editorial · [2 sources] ·

Researchers have developed a new availability backdoor attack called Event Burst Trigger (EBT) specifically for event-based Spiking Neural Networks (SNNs) used in object detection. This attack injects triggers into training data that cause temporally concentrated event streams during inference, significantly increasing the computational load on post-processing stages like Non-Maximum Suppression (NMS). While EBT largely preserves detection accuracy, it can increase NMS latency by up to 38%, potentially making it a bottleneck. The attack also subtly elevates resource utilization on edge platforms without obvious spikes, and standard detection methods like STRIP struggle to identify it. AI

IMPACT This research highlights a new vulnerability in event-based SNNs, potentially impacting the security and reliability of edge AI systems.

RANK_REASON The cluster contains a research paper detailing a novel attack method on a specific type of neural network.

Read on arXiv cs.CV →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

New attack targets event-based SNNs, increasing latency by 38%

COVERAGE [2]

  1. arXiv cs.CV TIER_1 English(EN) · Jaesun Baek, Chanwook Lee, Eun-Kyu Lee ·

    Event Burst Trigger: An Availability Backdoor Attack on Event-Based SNN Object Detection

    arXiv:2607.09115v1 Announce Type: new Abstract: Event-based vision and spiking neural networks (SNNs) are increasingly adopted for edge intelligence under strict latency and energy constraints. However, the vulnerability of event-based SNN object detection models to availability …

  2. arXiv cs.CV TIER_1 English(EN) · Eun-Kyu Lee ·

    Event Burst Trigger: An Availability Backdoor Attack on Event-Based SNN Object Detection

    Event-based vision and spiking neural networks (SNNs) are increasingly adopted for edge intelligence under strict latency and energy constraints. However, the vulnerability of event-based SNN object detection models to availability backdoor attacks remains insufficiently studied.…

RELATED ENTITIES

RELATED TOPICS