Izhikevich-Inspired Temporal Dynamics for Enhancing Privacy, Efficiency, and Transferability in Spiking Neural Networks

2025-05-07

AI Summary

Addressing the challenge of jointly optimizing privacy preservation, energy efficiency, and generalization in spiking neural networks (SNNs), this paper introduces two lightweight, Izhikevich-inspired temporal spike transformations at the input level: Poisson-Burst and Delayed-Burst. These methods decouple complex biophysical dynamics into trainable probabilistic encoding modules, seamlessly integrated into standard leaky integrate-and-fire (LIF) networks. The work systematically elucidates how temporal spike coding governs robustness against membership inference attacks (MIAs), enabling explicit, three-way trade-off design among privacy, accuracy, and energy efficiency. Evaluations on CIFAR-10/100 and TinyImageNet demonstrate that Poisson-Burst achieves state-of-the-art (SOTA) accuracy while reducing inference energy consumption by 32% and lowering MIA success rate from 48.7% to 12.4%. Delayed-Burst further suppresses the MIA success rate to 5.1%, with only a 1.3% accuracy drop.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Privacy

Application Category

• Security and Privacy: Data transparency and provenance
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Biological neurons exhibit diverse temporal spike patterns, which are believed to support efficient, robust, and adaptive neural information processing. While models such as Izhikevich can replicate a wide range of these firing dynamics, their complexity poses challenges for directly integrating them into scalable spiking neural networks (SNN) training pipelines. In this work, we propose two probabilistically driven, input-level temporal spike transformations: Poisson-Burst and Delayed-Burst that introduce biologically inspired temporal variability directly into standard Leaky Integrate-and-Fire (LIF) neurons. This enables scalable training and systematic evaluation of how spike timing dynamics affect privacy, generalization, and learning performance. Poisson-Burst modulates burst occurrence based on input intensity, while Delayed-Burst encodes input strength through burst onset timing. Through extensive experiments across multiple benchmarks, we demonstrate that Poisson-Burst maintains competitive accuracy and lower resource overhead while exhibiting enhanced privacy robustness against membership inference attacks, whereas Delayed-Burst provides stronger privacy protection at a modest accuracy trade-off. These findings highlight the potential of biologically grounded temporal spike dynamics in improving the privacy, generalization and biological plausibility of neuromorphic learning systems.

Problem

• Enhancing privacy and efficiency in Spiking Neural Networks
• Integrating Izhikevich-inspired temporal dynamics into scalable SNNs
• Evaluating spike timing effects on privacy and learning performance

Innovation

• Izhikevich-inspired temporal dynamics for SNNs
• Poisson-Burst and Delayed-Burst spike transformations
• Enhanced privacy and efficiency in neuromorphic systems