Authors: Gorka Abad, Stjepan Picek, Aitor Urbieta

Published on: February 05, 2024

Impact Score: 8.35

Arxiv code: Arxiv:2402.02886

Summary

• What is new: Development of a novel attack strategy tailored to spiking neural networks (SNNs) and federated learning (FL) that enhances attack effectiveness.
• Why this is important: Vulnerability of spiking neural networks and federated learning on low-powered devices to backdoor attacks using neuromorphic data.
• What the research proposes: A novel attack strategy that involves distributing the backdoor trigger temporally and across malicious devices.
• Results: Achievement of a 100% attack success rate, 0.13 MSE, and 98.9% SSIM.

Technical Details

Technological frameworks used: Federated learning, Spiking neural networks

Models used: Backdoor attack models tailored to SNNs and FL

Data used: Neuromorphic data

Potential Impact

Companies deploying low-powered devices using SNNs and FL, particularly in security-sensitive areas.

Want to implement this idea in a business?

We have generated a startup concept here: Neuroshield AI.