Semantic-aware Vehicular Edge Computing (VEC) has emerged as a novel paradigm to significantly reduce transmission costs and edge resource consumption by offloading extracted task-driven semantic information. However, excessive semantic extraction may impose additional computational workload. In the face of unknown environmental dynamics, the semantic extraction ratio must be jointly designed with task offloading for resource-constrained VEC. To this end, we conceive a multiple-objective (MO) semantic-aware task offloading framework for VEC by jointly optimizing semantic extraction ratio, transmit power and task scheduling strategies aimed at minimizing both long-term age-of-information (AoI) and energy consumption while guaranteeing queue stability. Subsequently, we propose a Lyapunov-guided multi-objective reinforcement learning (MORL)-based semantic-aware joint task scheduling and resource allocation (SJTSRA) solution. Specifically, Lyapunov optimization method is first leveraged to transform the original problem into a multi-objective Markov decision process (MOMDP). Then, the concave-augmented Pareto Q-learning (CAPQL) algorithm is employed to find Pareto optimal solutions through introducing uniform weight sampling and entropy regularization, where the convergence can be guaranteed theoretically. Simulation results show that the proposed solution achieves the closest approximation to the Pareto front with the highest hypervolume, and superior energy-AoI trade-offs across varying environment parameters compared to all benchmarks.

Large Language Models (LLMs) have shown remarkable capabilities across diverse applications, enabling broad user interaction. Despite advances in AI alignment, their susceptibility to adversarial jailbreak attacks presents serious safety challenges. To mitigate these risks, we propose MAJD, a Multi-Agent Jailbreak Defense Framework that operates entirely at inference time without retraining. Each agent performs specialized tasks: the Coordinator routes inputs based on semantic ambiguity, the Mutation Agent compresses prompts with predefined rules, the Analyzer conducts intent analysis and response generation, and the Revalidation Agent performs secondary response classification. By incorporating rulebased mutation and intent-based reasoning, our framework identifies the core malicious intent in both single-turn and multi-turn jailbreak attempts. Experimental results across multiple attack methods (AutoDAN, GPTFuzzer, PAIR, and MHJ) and across multiple target models demonstrate that MAJD significantly improves detection success rates. These results validate the efficiency and generalizability of MAJD as a scalable jailbreak defense solution for LLM applications.

Deep neural networks (DNNs) and natural language processing (NLP) systems have developed rapidly and have been widely used in various real-world fields. However, they have been shown to be vulnerable to backdoor attacks. Specifically, the adversary injects a backdoor into the model during the training phase, so that input samples with backdoor triggers are classified as the target class. Some attacks have achieved high attack success rates on the pre-trained language models (LMs), but there have yet to be effective defense methods. In this work, we propose a defense method based on deep model mutation testing. Our main justification is that backdoor samples are much more robust than clean samples if we impose random mutations on the LMs and that backdoors are generalizable. We first confirm the effectiveness of model mutation testing in detecting backdoor samples and select the most appropriate number of mutants and mutation operators. We then systematically defend against three extensively studied backdoor attack levels (i.e., char-level, word-level, and sentence-level) by detecting backdoor samples. We also make the first attempt to defend against the latest style-level backdoor attacks. We evaluate our approach on three benchmark datasets (i.e., IMDB, Yelp, and AG news) and three style transfer datasets (i.e., SST-2, Hate-speech, and AG news). The extensive experimental results demonstrate that our approach can detect backdoor samples more efficiently and accurately than the three state-of-the-art defense approaches.