Recently, deep unsupervised hashing has gained considerable attention in image retrieval due to its advantages in cost-free data labeling, computational efficiency, and storage savings. Although existing methods achieve promising performance by leveraging inherent visual structures within the data, they primarily focus on learning discriminative features from unlabeled images through limited internal knowledge, resulting in an intrinsic upper bound on their performance. To break through this intrinsic limitation, we propose a novel method, called Deep Unsupervised Hashing with External Guidance (DUH-EG), which incorporates external textual knowledge as semantic guidance to enhance discrete representation learning. Specifically, our DUH-EG: i) selects representative semantic nouns from an external textual database by minimizing their redundancy, then matches images with them to extract more discriminative external features; and ii) presents a novel bidirectional contrastive learning mechanism to maximize agreement between hash codes in internal and external spaces, thereby capturing discrimination from both external and intrinsic structures in Hamming space. Extensive experiments on four benchmark datasets demonstrate that our DUH-EG remarkably outperforms existing state-of-the-art hashing methods.

Fuzzing network protocols is challenging due to their various factors including protocol state and state transitions. To achieve better state coverage when fuzzing network protocol with grey-box fuzzing, several approaches are proposed to select valuable states and optimize the fuzzing process. Based on the ability of extensive knowledge integration and reasoning, large language models (LLMs) are also imported to generate more effective test cases for protocol fuzzing. However, these approaches leave poor state and code coverage on real-world services protocol evaluation since they use either random selection or heuristics. To address this issue, we present LLMgSSA, a Large Language Model guided State Selection Approach, which navigates the LLM to reason about protocol state selection. In the approach, LLMgSSA first extracts the features of the current protocol and state space and determines valuable states by interacting with the LLM. It then collects and analyzes the current status of each covered state and combines the inference results of the LLM for the final state selection. To evaluate the effectiveness of LLMgSSA, we have conducted extensive experiments with five real-world protocols from ProFuzzBench. Experimental results show that, compared to three state-of-the-art fuzzers, ChataFL, AFLnet, and NSFuzz, LLMgSSA can increase state transitions, covered states, branch coverage, and line coverage by up to 70.6%, 35.3%, 13.1%, and 13.1% respectively.

Code search aims to retrieve relevant code snippets from large code repositories based on query, promoting code reuse and enhancing software development efficiency.Deep Learning is a powerful approach for code search, in which the hard negative samples within training batches critically impact model performance.However, most existing deep code search models only randomly sample negative samples, resulting in a paucity or complete lack of hard negative samples.To address this limitation, we introduce a novel enhancement framework named HACS to optimize the composition of negative samples within batches, thus enhancing the training effectiveness of deep code search models.The core idea is to increase the count of hard negative samples within the negative samples corresponding to each query in the training batch.Specifically, HACS utilizes deep reinforcement learning techniques for sampling hard negative samples and implements vector-level mixed data augmentation strategy to generate hard negative samples.We evaluated our framework on a public dataset covering six programming languages.Experimental results reveal that HACS significantly improves the code search performance of existing models.