Poster
Fight Back Against Jailbreaking via Prompt Adversarial Tuning
Yichuan Mo · Yuji Wang · Zeming Wei · Yisen Wang
East Exhibit Hall A-C #4311
While Large Language Models (LLMs) have achieved tremendous success in various applications, they are also susceptible to jailbreak attacks. Several primary defense strategies have been proposed to protect LLMs from producing harmful information, mostly with a particular focus on harmful content filtering or heuristical defensive prompt designs. However, how to achieve intrinsic robustness through the prompts remains an open problem. In this paper, motivated by adversarial training paradigms for achieving reliable robustness, we propose an approach named Prompt Adversarial Tuning (PAT) that trains a prompt control attached to the user prompt as a guard prefix. To achieve our defense goal whilst maintaining natural performance, we optimize the control prompt with both adversarial and benign prompts. Comprehensive experiments show that our method is effective against both black-box and white-box attacks, reducing the success rate of advanced attacks to nearly 0 while maintaining the model's utility on the benign task. The proposed defense strategy incurs only negligible computational overhead, charting a new perspective for future explorations in LLM security.
Live content is unavailable. Log in and register to view live content