SpikingJET: Enhancing Fault Injection for Fully and Convolutional Spiking Neural Networks
Anil Bayram Gogebakan, Enrico Magliano, Alessio Carpegna, Annachiara Ruospo, Alessandro Savino, Stefano Di Carlo
TL;DR
SpikingJET tackles the challenge of evaluating the reliability of Spiking Neural Networks under hardware faults during inference. It introduces a simulation-based, non-intrusive fault injector that can inject stuck-at faults across five injection points (weights, membrane decay, threshold, membrane potential, and output spike) and three fault models, within an extensible framework built on open-source SNN tooling. Through experiments on N-MNIST, SHD, and DVS128, the approach demonstrates that most faults are masked and that vulnerability tends to concentrate in specific layers and parameters, notably those affecting spike timing and thresholds; this highlights where fault-tolerance strategies should be focused. The work provides a practical, scalable method for rapid resilience assessment of SNN-based systems in safety-critical domains, enabling agile exploration of fault scenarios and guiding reliability enhancements.
Abstract
As artificial neural networks become increasingly integrated into safety-critical systems such as autonomous vehicles, devices for medical diagnosis, and industrial automation, ensuring their reliability in the face of random hardware faults becomes paramount. This paper introduces SpikingJET, a novel fault injector designed specifically for fully connected and convolutional Spiking Neural Networks (SNNs). Our work underscores the critical need to evaluate the resilience of SNNs to hardware faults, considering their growing prominence in real-world applications. SpikingJET provides a comprehensive platform for assessing the resilience of SNNs by inducing errors and injecting faults into critical components such as synaptic weights, neuron model parameters, internal states, and activation functions. This paper demonstrates the effectiveness of Spiking-JET through extensive software-level experiments on various SNN architectures, revealing insights into their vulnerability and resilience to hardware faults. Moreover, highlighting the importance of fault resilience in SNNs contributes to the ongoing effort to enhance the reliability and safety of Neural Network (NN)-powered systems in diverse domains.