Towards Efficient Formal Verification of Spiking Neural Network (2408.10900v1)

Abstract: Recently, AI research has primarily focused on LLMs, and increasing accuracy often involves scaling up and consuming more power. The power consumption of AI has become a significant societal issue; in this context, spiking neural networks (SNNs) offer a promising solution. SNNs operate event-driven, like the human brain, and compress information temporally. These characteristics allow SNNs to significantly reduce power consumption compared to perceptron-based artificial neural networks (ANNs), highlighting them as a next-generation neural network technology. However, societal concerns regarding AI go beyond power consumption, with the reliability of AI models being a global issue. For instance, adversarial attacks on AI models are a well-studied problem in the context of traditional neural networks. Despite their importance, the stability and property verification of SNNs remains in the early stages of research. Most SNN verification methods are time-consuming and barely scalable, making practical applications challenging. In this paper, we introduce temporal encoding to achieve practical performance in verifying the adversarial robustness of SNNs. We conduct a theoretical analysis of this approach and demonstrate its success in verifying SNNs at previously unmanageable scales. Our contribution advances SNN verification to a practical level, facilitating the safer application of SNNs.