Sponge Examples Attack: Spikes Neural Network Energy Consumption by 100x
alexbilz · x · 2026-08-14
The paper "Sponge Examples: Energy-Latency Attacks on Neural Networks" introduces a Denial-of-Service (DoS) attack method targeting neural networks. By crafting specific input data, attackers can force the model to consume significantly more time and computational resources during inference.
Experiments show that this attack can increase the energy consumption of neural networks by 10x to 200x. In latency-critical applications like autonomous driving, such attacks could pose severe safety risks.
More from Safety
- 'The AI-driven cybersecurity black swan is happening' — and a reminder to enable 2FA — danielrock · 2026-10-16
- No-instruction autonomous agent wakes up, sets own goals, and tries hacking by default — sethlazar · 2026-10-16
- India's CERT-In says AI agents escaped testing environments and breached Indian organizations undetected for days — HimanshiET · 2026-10-16
- HTTP response headers can carry prompt injections against AI agents, researcher shows — arpit_bhayani · 2026-10-16
- OpenAI agent uses someone else's Ethereum testnet wallet to pay for its task — nptacek · 2026-10-16
- Interpretability won't save us: critics call superintelligence project reckless regardless — zetalyrae · 2026-10-16