Adversarial attacks on spiking convolutional neural networks for event-based vision.

• DOI: 10.5167/uzh-231251
• Publication Date: 2022-12-22
• Authors: Julian Büchel; Gregor Lenz; Yalun Hu; Sadique Sheik; Martino Sorbaro; Martino Sorbaro
• Keywords: FOS: Computer and information sciences; General Neuroscience; Computer Vision and Pattern Recognition (cs.CV); Computer Science - Computer Vision and Pattern Recognition; 2800 General Neuroscience; Neurosciences. Biological psychiatry. Neuropsychiatry; 02 engineering and technology; robust AI; spiking convolutional neural networks; adversarial examples; neuromorphic engineering; robust AI; dynamic vision sensors; adversarial examples; dynamic vision sensors; 0202 electrical engineering, electronic engineering, information engineering; 570 Life sciences; biology; spiking convolutional neural networks; neuromorphic engineering; 10194 Institute of Neuroinformatics; RC321-571; Neuroscience

Event-based dynamic vision sensors provide very sparse output in the form of spikes, which makes them suitable for low-power applications. Convolutional spiking neural networks model such event-based data and develop their full energy-saving potential when deployed on asynchronous neuromorphic hardware. Event-based vision being a nascent field, the sensitivity of spiking neural networks to potentially malicious adversarial attacks has received little attention so far. We show how white-box adversarial attack algorithms can be adapted to the discrete and sparse nature of event-based visual data, and demonstrate smaller perturbation magnitudes at higher success rates than the current state-of-the-art algorithms. For the first time, we also verify the effectiveness of these perturbations directly on neuromorphic hardware. Finally, we discuss the properties of the resulting perturbations, the effect of adversarial training as a defense strategy, and future directions.