Data poisoning-based backdoor attacks against supervised learning rules of Spiking Neural Networks

Key Points

• Backdoor attacks can compromise supervised learning, undermining the integrity of spiking neural networks.
• The analysis identifies vulnerabilities, permitting intrusions that can distort learning outputs.
• Using techniques like data poisoning, the methodology reveals weaknesses in machine learning frameworks.
• These findings highlight the need for better defenses against backdoor attacks in neural networks.