SafeLLM: Unlearning Harmful Outputs from Large Language Models against Jailbreak Attacks

Jailbreak attacks pose a serious threat to the safety of Large Language Models (LLMs) by crafting adversarial prompts that bypass alignment mechanisms, causing the models to produce harmful, restricted, or biased content. In this paper, we propose SafeLLM, a novel unlearning-based defense framework that unlearn the harmful knowledge from LLMs while preserving linguistic fluency and general capabilities. SafeLLM employs a three-stage pipeline: (1) dynamic unsafe output detection using a hybrid approach that integrates external classifiers with model-internal evaluations; (2) token-level harmful content tracing through feedforward network (FFN) activations to localize harmful knowledge; and (3) constrained optimization to suppress unsafe behavior without degrading overall model quality. SafeLLM achieves targeted and irreversible forgetting by identifying and neutralizing FFN substructures responsible for harmful generation pathways. Extensive experiments on prominent LLMs (Vicuna, LLaMA, and GPT-J) across multiple jailbreak benchmarks show that SafeLLM substantially reduces attack success rates while maintaining high general-purpose performance. Compared to standard defense methods such as supervised fine-tuning and direct preference optimization, SafeLLM offers stronger safety guarantees, more precise control over harmful behavior, and greater robustness to unseen attacks. Moreover, SafeLLM maintains the general performance after the harmful knowledge unlearned. These results highlight unlearning as a promising direction for scalable and effective LLM safety.

Bypassing DARCY Defense: Indistinguishable Universal Adversarial Triggers

Neural networks (NN) classification models for Natural Language Processing (NLP) are vulnerable to the Universal Adversarial Triggers (UAT) attack that triggers a model to produce a specific prediction for any input. DARCY borrows the "honeypot" concept to bait multiple trapdoors, effectively detecting the adversarial examples generated by UAT. Unfortunately, we find a new UAT generation method, called IndisUAT, which produces triggers (i.e., tokens) and uses them to craft adversarial examples whose feature distribution is indistinguishable from that of the benign examples in a randomly-chosen category at the detection layer of DARCY. The produced adversarial examples incur the maximal loss of predicting results in the DARCY-protected models. Meanwhile, the produced triggers are effective in black-box models for text generation, text inference, and reading comprehension. Finally, the evaluation results under NN models for NLP tasks indicate that the IndisUAT method can effectively circumvent DARCY and penetrate other defenses. For example, IndisUAT can reduce the true positive rate of DARCY's detection by at least 40.8% and 90.6%, and drop the accuracy by at least 33.3% and 51.6% in the RNN and CNN models, respectively. IndisUAT reduces the accuracy of the BERT's adversarial defense model by at least 34.0%, and makes the GPT-2 language model spew racist outputs even when conditioned on non-racial context.

Manipulating Predictions over Discrete Inputs in Machine Teaching

Machine teaching often involves the creation of an optimal (typically minimal) dataset to help a model (referred to as the `student') achieve specific goals given by a teacher. While abundant in the continuous domain, the studies on the effectiveness of machine teaching in the discrete domain are relatively limited. This paper focuses on machine teaching in the discrete domain, specifically on manipulating student models' predictions based on the goals of teachers via changing the training data efficiently. We formulate this task as a combinatorial optimization problem and solve it by proposing an iterative searching algorithm. Our algorithm demonstrates significant numerical merit in the scenarios where a teacher attempts at correcting erroneous predictions to improve the student's models, or maliciously manipulating the model to misclassify some specific samples to the target class aligned with his personal profits. Experimental results show that our proposed algorithm can have superior performance in effectively and efficiently manipulating the predictions of the model, surpassing conventional baselines.

Unveiling Security, Privacy, and Ethical Concerns of ChatGPT

This paper delves into the realm of ChatGPT, an AI-powered chatbot that utilizes topic modeling and reinforcement learning to generate natural responses. Although ChatGPT holds immense promise across various industries, such as customer service, education, mental health treatment, personal productivity, and content creation, it is essential to address its security, privacy, and ethical implications. By exploring the upgrade path from GPT-1 to GPT-4, discussing the model's features, limitations, and potential applications, this study aims to shed light on the potential risks of integrating ChatGPT into our daily lives. Focusing on security, privacy, and ethics issues, we highlight the challenges these concerns pose for widespread adoption. Finally, we analyze the open problems in these areas, calling for concerted efforts to ensure the development of secure and ethically sound large language models.