# Random heterogeneous spiking neural network for adversarial defense

## TL;DR

This paper introduces RandHet-SNN, a spiking neural network that improves robustness against adversarial attacks by mimicking biological neuron diversity.

## Contribution

The novel contribution is introducing randomized time decay constants to create neuron-level diversity in spiking neural networks for adversarial defense.

## Key findings

- RandHet-SNN significantly enhances adversarial robustness while preserving clean accuracy.
- The method outperforms randomization techniques in artificial neural networks.
- Neuronal heterogeneity and spiking variability improve network resilience.

## Abstract

Spiking neural networks (SNNs) offer a biologically inspired alternative to artificial neural networks (ANNs) by mimicking neuronal information transmission mechanisms. However, similar to ANNs, SNNs remain susceptible to adversarial examples, raising concerns about their robustness in practical applications. To address this vulnerability, we propose the Random Heterogeneous Spiking Neural Network (RandHet-SNN), inspired by the heterogeneity and stochasticity observed in biological neural systems. This architecture strengthens the network’s defense against adversarial attacks by introducing neuron-level diversity through randomized time decay constants, allowing each neuron to acquire unique temporal properties at every forward pass. We evaluate RandHet-SNN’s performance through extensive experiments with various adversarial attacks. Results indicate that RandHet-SNN significantly enhances network robustness while maintaining minimal impact on clean accuracy. RandHet-SNN displays significant potential for robust, energy-efficient neural computing in adversarial environments.

•Incorporating neuronal heterogeneity and trial-to-trial spiking variability•Random time constant significantly enhances adversarial robustness•RandHet-SNN outperforms randomization methods in ANN

Incorporating neuronal heterogeneity and trial-to-trial spiking variability

Random time constant significantly enhances adversarial robustness

RandHet-SNN outperforms randomization methods in ANN

Physics; Computer science; Engineering