LoopLLM: Transferable Energy-Latency Attacks in LLMs via Repetitive Generation
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917072957603840 |
|---|---|
| author | Li, Xingyu Liu, Xiaolei Liu, Cheng Xu, Yixiao Ding, Kangyi Xin, Bangzhou Yin, Jia-Li |
| author_facet | Li, Xingyu Liu, Xiaolei Liu, Cheng Xu, Yixiao Ding, Kangyi Xin, Bangzhou Yin, Jia-Li |
| contents | As large language models (LLMs) scale, their inference incurs substantial computational resources, exposing them to energy-latency attacks, where crafted prompts induce high energy and latency cost. Existing attack methods aim to prolong output by delaying the generation of termination symbols. However, as the output grows longer, controlling the termination symbols through input becomes difficult, making these methods less effective. Therefore, we propose LoopLLM, an energy-latency attack framework based on the observation that repetitive generation can trigger low-entropy decoding loops, reliably compelling LLMs to generate until their output limits. LoopLLM introduces (1) a repetition-inducing prompt optimization that exploits autoregressive vulnerabilities to induce repetitive generation, and (2) a token-aligned ensemble optimization that aggregates gradients to improve cross-model transferability. Extensive experiments on 12 open-source and 2 commercial LLMs show that LoopLLM significantly outperforms existing methods, achieving over 90% of the maximum output length, compared to 20% for baselines, and improving transferability by around 40% to DeepSeek-V3 and Gemini 2.5 Flash. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_07876 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | LoopLLM: Transferable Energy-Latency Attacks in LLMs via Repetitive Generation Li, Xingyu Liu, Xiaolei Liu, Cheng Xu, Yixiao Ding, Kangyi Xin, Bangzhou Yin, Jia-Li Cryptography and Security Artificial Intelligence Computation and Language As large language models (LLMs) scale, their inference incurs substantial computational resources, exposing them to energy-latency attacks, where crafted prompts induce high energy and latency cost. Existing attack methods aim to prolong output by delaying the generation of termination symbols. However, as the output grows longer, controlling the termination symbols through input becomes difficult, making these methods less effective. Therefore, we propose LoopLLM, an energy-latency attack framework based on the observation that repetitive generation can trigger low-entropy decoding loops, reliably compelling LLMs to generate until their output limits. LoopLLM introduces (1) a repetition-inducing prompt optimization that exploits autoregressive vulnerabilities to induce repetitive generation, and (2) a token-aligned ensemble optimization that aggregates gradients to improve cross-model transferability. Extensive experiments on 12 open-source and 2 commercial LLMs show that LoopLLM significantly outperforms existing methods, achieving over 90% of the maximum output length, compared to 20% for baselines, and improving transferability by around 40% to DeepSeek-V3 and Gemini 2.5 Flash. |
| title | LoopLLM: Transferable Energy-Latency Attacks in LLMs via Repetitive Generation |
| topic | Cryptography and Security Artificial Intelligence Computation and Language |
| url | https://arxiv.org/abs/2511.07876 |