Lingzhi Li’s research while affiliated with Soochow University (PRC) and other places

In intrusion detection systems, deep learning has demonstrated its capability to effectively mine flow representations, significantly enhancing the ability to detect anomalies. However, current approaches still suffer from limitations in flow feature extraction and may require fine-tuning on different forms of data, and may even be nontransferable. The task of accurately and efficiently handling multiple forms of flow remains a challenging endeavor. In this work, we propose the Hypergraph Recurrent Neural Network (HRNN), a novel intrusion detection method that leverages the hypergraph higher-order structure and recurrent network. We construct flow data as hypergraph structures, which allow for more abundant information representation and implicitly incorporate more similar information in the model. The recurrent module extracts temporal features of the flow. Our design effectively fuses representations imbued with rich spatial and temporal semantics. Evaluations of several publicly available datasets portray that HRNN outperforms other state-of-the-art methods.

Coded computing has proven its efficiency in tolerating stragglers in distributed computing. Workers return the sub-computation results to the master after computing, and the master recovers the final computation result by decoding. However, the workers may provide incorrect results, which leads to wrong final result. Therefore, it is meaningful to improve the resilience of coded computing against errors. Most existing verification schemes only use the workers’ fully correct computations to recover the final result, and the defective computations are not considered for decoding. In this paper, we focus on matrix multiplication and design a general Test-and-Decode (TD) scheme to recover the final result efficiently. Furthermore, we divide each sub-computation result into multiple parts and fully use the correct parts for partial recovery, which can improve the tolerance for errors in computations. Decoding is performed only when the verification result satisfies the permission, which avoids repetitive decoding. We conduct extensive simulation experiments to evaluate the probability of successful recovery of the results and the computation time of the TD scheme. We also compare the TD scheme with other verification schemes and the results show that it outperforms the current schemes in terms of efficiency in verifying and recovering computational results.

Recently, permissioned blockchain has been extensively explored in various fields, such as asset management, supply chain, healthcare, and many others. Many scholars are dedicated to improving its verifiability, scalability, and performance based on sharding techniques, including grouping nodes and handling cross-shard transactions. However, they ignore the node vulnerability problem, i.e., there is no guarantee that nodes will not be maliciously controlled throughout their life cycle. Facing this challenge, we propose RecAGT, a novel identification scheme aimed at reducing communication overhead and identifying potential malicious nodes. First, shard testable codes are designed to encode the original data in case of a leak of confidential data. Second, a new identity proof protocol is presented as evidence against malicious behavior. Finally, adaptive group testing is chosen to identify malicious nodes. Notably, our work focuses on the internal operation within the committee and can thus be applied to any sharding permissioned blockchains. Simulation results show that our proposed scheme can effectively identify malicious nodes with low communication and computational costs.

Traffic congestion, air pollution, fuel wastage, and car accidents are all exacerbated by increased traffic. Thus, vehicular communications, which refer to information transmission between cars, pedestrians, and infrastructures, have lately gained popularity and been extensively explored due to their enormous potential to enable intelligent transportation and various safety applications. Manually piloted cars and automated vehicles can acquire relevant information via vehicular communications to enhance traffic security and boost entertainment services. The basic concept of automobile clouds was originally published in the literature not long ago, and several suggested structural approaches have been presented in this study thus far. Several academics have concentrated on the structural layout to address various problems and, as a result, satisfy user expectations in order to give dependable services. We examined various vehicular cloud topologies in this study. We also offered a complete summary of current network layer research on allowing efficient vehicle communications and examined specific security, architectural, and reliability concerns in vehicular clouds. Also, the taxonomy of vehicular networks was discussed in terms of the service link between vehicular networks and cloud computing. Ultimately, we discussed the research prospects available. The results showed that security and privacy challenges are among the most important challenges.

Edge computing is an efficient computing paradigm, which can utilize computing devices at the edge of network to provide real-time proximity service. Since edge devices lack centralized management, they are more vulnerable to being attacked. Therefore, the issues of data security and user privacy in edge computing are particularly important. A large number of existing literature focus on the data security and user privacy with independent attackers. However, cooperative attacks, in which multiple attackers can collaborate to obtain the data content and user privacy, have not been fully investigated. In particular, we take the matrix-vector multiplication which is a basic component of most machine learning algorithms as the basic task. Therefore, in this paper, we focus on the Secure and Privacy Matrix-vector Multiplication (SPMM) issue for edge computing against cooperative attack and design a general coded computation scheme to achieve lowest system resource consumption, i.e. communication cost and computational load. Specifically, we propose two coding schemes: Secure and Private Coding with lower communication Cost (SPCC) and Secure and Private Coding with lower computational Load (SPCL). We also conduct solid theoretical analyses and extensive experiments to demonstrate that both two proposed coding schemes can achieve lower communication cost and computational load than existing work. Finally, we perform extensive analyses to the superiority of the proposed schemes.

Compliance with security requirements in Wireless Sensor Networks (WSNs) is known as an important phenomenon in conserving energy consumption due to its dynamic topology. Energy consumption management in WSN can be achieved through secure routing protocols. Authentication, firewall and intrusion detection, and trust management are common ways to provide security on the WSNs. Meanwhile, trust management provides better performance due to ensuring distributed security through collaboration. In this paper, a Trust-Aware Dynamic Routing algorithm based on Extended AODV protocol for secure communications in the WSN (TADR-EAODV) is proposed. TADR-EAODV considers factors such as direct trust, recommended trust, connectivity strength, energy rate and worthiness score to measure the distributed safety level of nodes in routing. We use the AODV protocol for routing work, which is extended through a multi-route routing approach. In addition, a centralized ensemble clustering for node grouping is included in TADR-EAODV, which enables clustering-based routing to enhance WSN performance. Extensive experiments have been performed in the presence of denial-of-service attacks to verify our claims. The experimental results prove that high-performance TADR-EAODV has been able to identify attackers by detecting their anomalous behavior. Besides, TADR-EAODV has improved the average packet transfer rate by 6% to 17% compared to existing state-of-the-art algorithms such as LEACH-TM, TAAO-SDTIM and TAOSC-MHR.

Deep learning provides many new and efficient solutions for edge computing. We study training image classification models on edge devices in this paper. Although there have been many researches on deep learning in edge computing. Most of them did not consider the impact of the limited service capabilities of edge devices, the problem of straggler and insecurity of training data on the system. We design a new distributed computing system to train image classification models on edge devices. To be more specific, we vectorize the convolutional neural network (CNN) to transform it to a lot of matrix multiplications. These matrix multiplications can be arbitrarily cut into many smaller matrix multiplications suitable for computing on edge devices. Besides, our system utilizes codes to ensure the stability and security of distributed matrix multiplications on edge devices. In the performance evaluation, we test the performance of matrix multiplications and a CNN model training in our system with uncoded and coded strategies. The evaluation results show that the system with code strategies perform better than with uncoded strategies on the edge devices having the problem of straggler. In summary, we design a secure distributed image classification model training system for heterogenous edge computing.