How can we enable models to comprehend video anomalies occurring over varying temporal scales and contexts? Traditional Video Anomaly Understanding (VAU) methods focus on frame-level anomaly prediction, often missing the interpretability of complex and diverse real-world anomalies. Recent multimodal approaches leverage visual and textual data but lack hierarchical annotations that capture both short-term and long-term anomalies. To address this challenge, we introduce HIVAU-70k, a large-scale benchmark for hierarchical video anomaly understanding across any granularity. We develop a semi-automated annotation engine that efficiently scales high-quality annotations by combining manual video segmentation with recursive free-text annotation using large language models (LLMs). This results in over 70,000 multi-granular annotations organized at clip-level, event-level, and video-level segments. For efficient anomaly detection in long videos, we propose the Anomaly-focused Temporal Sampler (ATS). ATS integrates an anomaly scorer with a density-aware sampler to adaptively select frames based on anomaly scores, ensuring that the multimodal LLM concentrates on anomaly-rich regions, which significantly enhances both efficiency and accuracy. Extensive experiments demonstrate that our hierarchical instruction data markedly improves anomaly comprehension. The integrated ATS and visual-language model outperform traditional methods in processing long videos. Our benchmark and model are publicly available at https://github.com/pipixin321/HolmesVAU.

In recent years, video anomaly detection has been extensively investigated in both unsupervised and weakly supervised settings to alleviate costly temporal labeling. Despite significant progress, these methods still suffer from unsatisfactory results such as numerous false alarms, primarily due to the absence of precise temporal anomaly annotation. In this paper, we present a novel labeling paradigm, termed "glance annotation", to achieve a better balance between anomaly detection accuracy and annotation cost. Specifically, glance annotation is a random frame within each abnormal event, which can be easily accessed and is cost-effective. To assess its effectiveness, we manually annotate the glance annotations for two standard video anomaly detection datasets: UCF-Crime and XD-Violence. Additionally, we propose a customized GlanceVAD method, that leverages gaussian kernels as the basic unit to compose the temporal anomaly distribution, enabling the learning of diverse and robust anomaly representations from the glance annotations. Through comprehensive analysis and experiments, we verify that the proposed labeling paradigm can achieve an excellent trade-off between annotation cost and model performance. Extensive experimental results also demonstrate the effectiveness of our GlanceVAD approach, which significantly outperforms existing advanced unsupervised and weakly supervised methods. Our annotations and code are publicly available at https://github.com/pipixin321/GlanceVAD.

The requirement for deploying deep learning (DL) models efficiently has boosted the research of DL compilers. Especially, the difficulty of generating optimized tensor programs has driven DL compilers to commonly adopt the auto-tuning approaches. Consequently, there are increasing demands to improve the effectiveness of auto-tuning in terms of both search efficiency and search quality. However, existing auto-tuning approaches commonly treat subgraphs individually and overlook the similarities among them, and thus fail to generate better tensor programs under limited time budget. To address the above drawbacks, we propose FamilySeer, an auto-tuning framework that can generate better tensor programs by exploiting the subgraph similarities. Specifically, FamilySeer organizes similar subgraphs into subgraph families, where the cost models are built at family basis with improved accuracy for estimating high potential program candidates. To further leverage the similarity, FamilySeer uses the accurate cost model per family to reduce the number of program candidates for costly hardware measurements without degrading search quality. The experiment results on various DL models demonstrate that FamilySeer can achieve better search efficiency/quality on both CPU and GPU platforms compared to the state-of-the-art auto-tuning framework.

Blockchain has proven in sensor networks as a distributed solution for transparent and secure storage, which allows its application in mobile wireless sensor networks (MWSNs). The consensus mechanism, an essential aspect of blockchain technology, must concern the high mobility, resource-constrained nature, and weak physical defenses of sensor nodes in MWSNs. To secure MWSN data storage in clustered communication, we design a proof-of-information (PoI) variant for fair miner campaigning via the amount of valid data generated from environmental information, including a dynamic adjustment of the data volume threshold to detect malicious nodes and prevent them from misreporting information. Additionally, we introduce a filtering mechanism through the dynamic integrated trust (DIt) of nodes, which integrates the trust evaluation of peer nodes across the network combining objective performance to prevent malicious nodes from infiltrating the final consensus group. The multi-level filtering technique improves the campaign fairness while isolating malicious nodes, ensuring complexity-sensitive PBFT algorithm efficiency in large-scale networks. Simulation results show that the scheme isolates 90% of the malicious nodes and screens 20% of members to produce a smaller final consensus group. Further analysis of impacts on the performance considering network topology and mobility patterns and comparisons of the relevant solutions are presented.

Reversible data hiding in encrypted images (RDH-EI) technology is commonly used in cloud storage images for privacy protection. Most existing RDH-EI techniques reported in the literature applied block permutation and co-modulation (BPCM) encryption to generate encrypted images. This work analyses the security of the RDH-EI algorithm based on BPCM encryption under known plaintext attacks (KPAs). Different from the existing KPAs, this paper considers that attackers can perform KPAs based on marked encrypted images and shows that BPCM encryption has the risk of information leakage. To find the constant features of a block before and after co-modulation, the first-pixel difference block (FDB) of a block is first defined. Then, a pseudo cypher difference image of the cyphertext image is constructed to eliminate the changed FDBs so that the differences in the cyphertext FDBs are the same as the FDBs in the corresponding plaintext difference image. Finally, we design an FDB-based block permutation key estimation method according to the plaintext difference image and pseudocyphertext difference image. The influence of block size on key estimation accuracy and the time complexity of the proposed KPA algorithm are analysed and discussed. Experimental results show that the correct rate of key estimation is positively correlated with the block size and the number of plain-cyphertext pairs. The average correct rate of key estimation reaches 63% when the block size is greater than 3×3.