Search

Social networks have become an indispensable part of contemporary life, and the widespread occurrence of cyberbullying among young people on these sites has sparked serious societal issues. Information on social media often encompasses various heterogeneous elements, such as locations, individuals, organizations, and more. Despite this complexity, text analysis has dominated most studies on cyberbullying detection, ignoring the variety of media data present in social networks. This paper proposes a novel framework for tackling this gap: RSBully is a multi-agent Reinforced Guided Weighted Multi-relational Subgraph Neural Network that is intended to detect cyberbullying effectively. It organizes elements of different types using heterogeneous information networks (HINs) and then transforms them into a weighted multi-relation graph to model the relationship between social media sessions. Second, RSBully uses a reinforced subgraph neural network to eliminate redundant information in social media data by extracting the prominent subgraphs of the graph and capturing intrinsic correlation between modalities and bullying behaviors with interpretability. Comprehensive experimental assessments on real-world social media datasets show that the performance of the RSBully framework outperforms various current state-of-the-art models. [ABSTRACT FROM AUTHOR]

The study of adversarial attacks is crucial in the design of robust neural network models. In this work, we propose a hyperparameter optimization framework for black-box adversarial attacks. We observe that hyperparameters are extremely important to enhance the query efficiency of many black-box adversarial attack methods. Hence, we propose an adaptive hyperparameter tuning framework such that, in each query iteration, the attacker can adaptively selects the hyperparameter configuration based on the feedback from the victim to improve the attack success rate and query efficiency of the attack algorithm. The experiment results show, by adaptively tuning the attack hyperparameters, our technique outperforms the original algorithm, where the query efficiency is improved by 33.63% on the NES algorithm for untargeted attacks, 44.47% on the Bandits algorithm for untargeted attacks, and 32.24% improvement on the Bandits algorithm for targeted attacks. [ABSTRACT FROM AUTHOR]

Efficient automated presswork defect detection is valuable to the printing industry since such defects dramatically depress the presswork grad and manually detecting them is cost-ineffective. Instead of using real defect annotating data, the goal of this paper is to just use defect-free high-resolution images to complete the detection of defect in presswork (SWDF). In this paper, we first propose an end-to-end cropping method to balance the number of samples which cropped from different patterns in the defect-free high-resolution images, and thus, we can solve the quantity imbalance between different patterns (image level). Secondly, we designed an simple but effective loss function to solve the severe imbalance which caused by the larger quantity difference between the defective pixels and background pixels (pixel level). In order to solve the absence of the real defect annotating data and replace the time-consuming acquisition of defect data, we designed a high-speed defect generation algorithm, which directly generate defect on defect-free samples (obtained from the cropping process). As for the detection of defects, we use self-attention to design a novel and effective semantic segmentation head (GST), which can exploit global information from the feature map to repair the detection results, so as to obtain a better performance. In the experimental part, we will use a presswork defect dataset from the Zhentu Cup Competition (ZCC) and a public dataset DAGM 2007 to test the effect of our scheme. Especially, compared with existing supervised models, our model has also reached the state of the art in DAGM 2007. [ABSTRACT FROM AUTHOR]

When users store data in big data platforms, the integrity of outsourced data is a major concern for data owners due to the lack of direct control over the data. However, the existing remote data auditing schemes for big data platforms are only applicable to static data. In order to verify the integrity of dynamic data in a Hadoop big data platform, we presents a dynamic auditing scheme meeting the special requirement of Hadoop. Concretely, a new data structure, namely Data Block Index Table, is designed to support dynamic data operations on HDFS (Hadoop distributed file system), including appending, inserting, deleting, and modifying. Then combined with the MapReduce framework, a dynamic auditing algorithm is designed to audit the data on HDFS concurrently. Analysis shows that the proposed scheme is secure enough to resist forge attack, replace attack and replay attack on big data platform. It is also efficient in both computation and communication. [ABSTRACT FROM AUTHOR]

An effective and secure system used for evidence preservation is essential to possess the properties of anti-loss, anti-forgery, anti-tamper and perfect verifiability. Traditional architecture which relies on centralized cloud storage is depressingly beset by the security problems such as incomplete confidence and unreliable regulation. Moreover, an expensive, inefficient and incompatible design impedes the effort of evidence preservation. In contrast, the decentralized blockchain network is qualified as a perfect replacement for its secure anonymity, irrevocable commitment, and transparent traceability. Combining with subliminal channels in blockchain, we have weaved the transaction network with newly designed evidence audit network. In this paper, we have presented and implemented a lightweight digital evidence-preservation architecture which possesses the features of privacy-anonymity, audit-transparency, function-scalability and operation-lightweight. The anonymity is naturally formed from the cryptographic design, since the cipher evidence under encrypted cryptosystem and hash-based functions leakages nothing to the public. Covert channels are efficiently excavated to optimize the cost, connectivity and security of the framework, transforming the great computation power of Bitcoin network to the value of credit. The transparency used for audit, which relates to the proof of existence, comes from instant timestamps and irreversible hash functions in mature blockchain network. The scalability is represented by the evidence chain interacted with the original blockchain, and the extended chains on top of mainchain will cover the most of auditors in different institutions. And the lightweight, which is equal to low-cost, is derived from our fine-grained hierarchical services. At last, analyses of efficiency, security, and availability have shown the complete accomplishment of our system. [ABSTRACT FROM AUTHOR]