Large language models (LLMs) can transform natural language instructions into executable commands for IoT devices like unmanned aerial vehicles (UAVs), creating new development opportunities. However, safety concerns about LLMs translating commands into machine or program control instructions cannot be overlooked. Currently, jailbreak instructions used to test the LLM security are often restricted to specific modes or tasks, resulting in a lack of diversity and leaving some tasks unexplored. To address this issue, we introduce a Multi-Instruction Fusion (MIF) method that can automatically fuse harmful prompts and various task instructions into jailbreaks. Firstly, we adopt a reverse decomposition strategy to acquire sufficient supervised data for fusing harmful prompts and instructions into jailbreaks and construct a task instruction synthesizer based on it. Then, to determine the optimal instruction combinations in the vast combination space, we propose a representative-node-based selection strategy, ReNB, to rank and filter the instruction combinations on a few representative samples, thereby accelerating the identification of the valid ones. Experimental results demonstrate that MIF significantly improves the attack success rate, achieving over 90% on GPT-4o-mini, LLaMa2-70B and Qwen2-7B models, outperforming the state-of-the-art baselines.

Multi-relational graph learning aims to embed entities and relations in knowledge graphs into low-dimensional representations, which has been successfully applied to various multi-relationship prediction tasks, such as information retrieval, question answering, and etc. Recently, contrastive learning has shown remarkable performance in multi-relational graph learning by data augmentation mechanisms to deal with highly sparse data. In this paper, we present a Multi-Relational Graph Contrastive Learning architecture (MRGCL) for multi-relational graph learning. More specifically, our MRGCL first proposes a Multi-relational Graph Hierarchical Attention Networks (MGHAN) to identify the importance between entities, which can learn the importance at different levels between entities for extracting the local graph dependency. Then, two graph augmented views with adaptive topology are automatically learned by the variant MGHAN, which can automatically adapt for different multi-relational graph datasets from diverse domains. Moreover, a subgraph contrastive loss is designed, which generates positives per anchor by calculating strongly connected subgraph embeddings of the anchor as the supervised signals. Comprehensive experiments on multi-relational datasets from three application domains indicate the superiority of our MRGCL over various state-of-the-art methods. Our datasets and source code are published at https://github.com/Legendary-L/MRGCL.

The importance of social network alignment (SNA) for various downstream applications, such as social network information fusion and e-commerce recommendation, has prompted numerous professionals to develop and share SNA tools. However, malicious actors can exploit these tools to integrate sensitive user information, thereby posing cybersecurity risks. Although many researchers have explored attacking SNA (ASNA) through network modification attacks to protect users, practical feasibility remains challenging. In this study, we propose an effective node injection attack via a dynamic programming framework (DPNIA) to address the problem of modeling and solving ASNA within a limited time and balancing the costs and benefits. DPNIA models ASNA as a problem of maximizing the number of confirmed incorrect correspondent node pairs with greater similarity scores than the pairs between existing nodes, thereby making ASNA solvable. A cross-network evaluation method is employed directly to identify node vulnerabilities, facilitating progressive attacking from easy to difficult. In addition, an optimal injection strategy searching method based on dynamic programming is used to determine which links should be added between the injected and existing nodes, thereby enhancing the effectiveness of the attack at a low cost. Experiments on four real-world datasets demonstrated that DPNIA consistently and significantly surpasses various baselines when attacking both multiple networks simultaneously and a single network.

REST API Fuzzing is an emerging approach for automated vulnerability detection in cloud services. However, existing SOTA fuzzers face challenges in generating lengthy sequences comprising high-semantic requests, so that they may hardly trigger hard-to-reach states within a cloud service. To overcome this problem, we propose RESTLess, a flexible and efficient approach with hybrid optimization strategies for REST API fuzzing enhancement. Specifically, to pass the cloud gateway syntax semantic checking, we construct a dataset of valid parameters of REST API with Large Language Model named RTSet, then utilize it to develop an efficient REST API specification semantic enhancement approach. To detect vulnerability hidden under complex API operations, we design a flexible parameter rendering order optimization algorithm to increase the length and type of request sequences. Evaluation results highlight that RESTLess manifests noteworthy enhancements in the semantic quality of generated sequences in comparison to existing tools, thereby augmenting their capabilities in detecting vulnerabilities effectively. We also apply RESTLess to nine real-world cloud service such as Microsoft Azure, Amazon Web Services, Google Cloud, etc., and detecte 38 vulnerabilities, of which 16 have been confirmed and fixed by the relevant vendors.

Online users are typically active on multiple social media networks (SMNs), which constitute a multiplex social network. With improvements in cybersecurity awareness, users increasingly choose different usernames and provide different profiles on different SMNs. Thus, it is becoming increasingly challenging to determine whether given accounts on different SMNs belong to the same user; this can be expressed as an interlayer link prediction problem in a multiplex network. To address the challenge of predicting interlayer links, feature or structure information is leveraged. Existing methods that use network embedding techniques to address this problem focus on learning a mapping function to unify all nodes into a common latent representation space for prediction; positional relationships between unmatched nodes and their common matched neighbors (CMNs) are not utilized. Furthermore, the layers are often modeled as unweighted graphs, ignoring the strengths of the relationships between nodes. To address these limitations, we propose a framework based on multiple types of consistency between embedding vectors (MulCEVs). In MulCEV, the traditional embedding-based method is applied to obtain the degree of consistency between the vectors representing the unmatched nodes, and a proposed distance consistency index based on the positions of nodes in each latent space provides additional clues for prediction. By associating these two types of consistency, the effective information in the latent spaces is fully utilized. In addition, MulCEV models the layers as weighted graphs to obtain representation. In this way, the higher the strength of the relationship between nodes, the more similar their embedding vectors in the latent representation space will be. The results of our experiments on several real-world and synthetic datasets demonstrate that the proposed MulCEV framework markedly outperforms current embedding-based methods, especially when the number of training iterations is small.

The classification problem of imbalanced datasets has received much attention in recent years. This imbalance problem usually occurs in when the ratio between classes is high. Many techniques have been developed to tackle the imbalance problem in supervised learning. The Synthetic Minority Over-sam pling Technique (SMOTE) is one of the most effective over-sampling methods processing this problem, which changes the distribution of training sets to balance the different number of examples of each class. However, SMOTE randomly synthesizes the minority instances along a line joining a minority instance and its selected nearest neighbors, ignoring nearby majority instances and isolated points, which would affect the final classification result. In this paper, we propose two improved techniques based on SMOTE through sparse representation theory. This extension results in Sparse-SMOTE and SROT (Sparse Representation Based Over-Sampling Technique). The Sparse-SMOTE replaces the k-nearest neighbors of the SMOTE with sparse representation, and the SROT uses a sparse dictionary to create a synthetic sample directly. The experiments are performed on 10 UCI datasets using C4.5 as the learning algorithm. The experimental results show that both proposed methods can achieve better performance on TP-Rate, F-Measure, G-Mean and AUC values. Moreover, the results show that our new proposals’ perform is more effective compared with SMOTE and some other approaches.