Large Language Models (LLMs) have demonstrated remarkable success across diverse applications, yet their susceptibility to malicious exploitation remains a critical challenge. Notably, LLMs are known to be vulnerable to jailbreaking attacks, where adversaries craft malicious inputs to induce harmful or unethical outputs. In this paper, motivated by the unique effectiveness and scalability of In-Context Learning (ICL) in LLMs, we explore its potential to modulate the safety alignment of LLMs. Specifically, we propose the In-Context Attack (ICA), which employs harmful demonstrations to subvert LLMs’ safety, and the In-Context Defense (ICD), which bolsters their resilience through examples that demonstrate refusal to produce harmful responses. By adjusting the distribution of safety in LLM outputs through adversarial demonstrations, our proposed in-context attack and defense facilitate effective manipulation of their alignment. We first provide theoretical insights to illustrate how minimal in-context demonstrations can efficiently alter safety alignment. Empirically, we validate ICA and ICD across multiple models, datasets, and attack baselines, showing their efficacy and scalability for red-teaming evaluations and robust safeguards for real-world deployment. Overall, our work unveils the pivotal yet understudied role of ICL in LLM safety, opening new avenues for understanding and improving them.

The security situation of the Internet of Things (IoT) is more serious than ever, and there is an urgent need to detect and patch device vulnerability rapidly. With the astronomical numbers of IoT devices, it is very difficult to execute regular security inspections. Existing vulnerability detection technology based on simple feature matching cannot reach high accuracy to detect firmware vulnerabilities, while using control flow graph matching directly has proven to be too expensive. To address the problem of accurate and efficient, we present a method of staged firmware vulnerability detection based on code similarity. First stage, function embedding based on neural network is used to analyze the similarities among functions, and large-scale firmware security inspection can be achieved efficiently. Second stage, the similarity among function local call flow graphs is calculated for fine-grained firmware security analysis, and this stage can improve the accuracy of vulnerability detection.We compared our method with state-of-the-art approaches, and the experimental results demonstrate that our method is more accurate. The average retraining time of our method is 1 h, and the real-world firmware vulnerability detection experiment of our method demonstrates that the true positive rate of the top 30 is as high as 86%.