ZZZhongwen Zhu

Zhongwen Zhu

*This is an algorithmically generated author record based on publicly available information.

Zhongwen Zhu has a total of 21 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Communication-Efficient Network Topology in Decentralized Learning: A Joint Design of Consensus Matrix and Resource Allocation

    by Jingrong Wang, Ben Liang, Zhongwen Zhu, Emmanuel Thepie Fapi, Hardik Dalal

    In decentralized machine learning over a network of workers, each worker updates its local model as a weighted average of its local model and all models received from its neighbors. Efficient consensus weight matrix design and communication resource allocation can increase the training convergence rate and reduce the wall-clock training time. In this paper, we jointly consider these two factors and propose a novel algorithm termed Communication-Efficient Network Topology (CENT), which reduces the latency in each training iteration by removing unnecessary communication links. CENT enforces communication graph sparsity by iteratively updating, with a fixed step size, a trade-off factor between the convergence factor and a weighted graph sparsity. We further extend CENT to one with an adaptive step size (CENT-A), which adjusts the trade-off factor based on the feedback of the objective function value, without introducing additional computation complexity. We show that both CENT and CENT-A preserve the training convergence rate while avoiding the selection of poor communication links. Numerical studies with real-world machine learning data in both homogeneous and heterogeneous scenarios demonstrate the efficacy of CENT and CENT-A and their performance advantage over state-of-the-art algorithms.


  • Open Access

    EtherShield: Time-interval Analysis for Detection of Malicious Behavior on Ethereum

    by Bofeng Pan, Natalia Stakhanova, Zhongwen Zhu

    Advances in blockchain technology have attracted significant attention across the world. The practical blockchain applications emerging in various domains, ranging from finance, healthcare, and entertainment, have quickly become attractive targets for adversaries. The novelty of the technology coupled with the high degree of anonymity it provides made malicious activities even less visible in the blockchain environment. This made their robust detection challenging. This article presents EtherShield, a novel approach for identifying malicious activity on the Ethereum blockchain. By combining temporal transaction information and contract code characteristics, EtherShield can detect various types of threats and provide insight into the behavior of contracts. The time-interval-based analysis used by EtherShield enables expedited detection, achieving comparable accuracy to other approaches with significantly less data. Our validation analysis, which involved over 15,000 Ethereum accounts, demonstrated that EtherShield can significantly expedite the detection of malicious activity while maintaining high accuracy levels (86.52% accuracy with 1 hour of transaction history data and 91.33% accuracy with 1 year of transaction history data).


  • Unsupervised GAN-Based Intrusion Detection System Using Temporal Convolutional Networks and Self-Attention

    by Paulo Freitas de Araujo-Filho, Mohamed Naili, Georges Kaddoum, Emmanuel Thepie Fapi, Zhongwen Zhu

    Fifth-generation (5G) networks provide connectivity to a massive number of devices and boost a plethora of applications in several different domains. However, the large adoption of connected devices increases attack surfaces and introduces several security threats that can severely damage physical objects and risk people’s lives. Despite existing intrusion detection systems (IDSs), there are still several challenges to be addressed in the detection of cyber-attacks. For instance, while unsupervised IDSs are required to detect zero-day attacks, they usually present high false positive rates. Moreover, most existing IDSs rely on long short-term memory (LSTM) networks to consider time-dependencies among data. However, LSTM networks have recently been shown to present several drawbacks and limitations, which put into question their performance on sequence modeling tasks. Thus, in this paper, we investigate generative adversarial networks (GANs), a promising unsupervised approach to detecting attacks by implicitly modeling systems, and alternatives to LSTM networks to consider temporal dependencies among data. We propose a novel unsupervised GAN-based IDS that uses temporal convolutional networks (TCNs) and self-attention to detect cyber-attacks. The proposed IDS leverages edge computing and is proposed for edge servers, which bring computation resources closer to end nodes. Experiment results show that our proposed IDS can be configured to satisfy different detection rate and detection time requirements. Moreover, they show that our IDS is more accurate and at least 3.8 times faster than two state-of-the-art GAN-based IDSs that are used as baselines.


  • Open Access

    Network Intrusion Detection via Flow-to-Image Conversion and Vision Transformer Classification

    by Chi Mai Kim Ho, Kin-Choong Yow, Zhongwen Zhu, Sarang Aravamuthan

    In recent years, computer networks have become an indispensable part of our life, and these networks are vulnerable to various type of network attacks, compromising the security of our data and the freedom of our communications. In this paper, we propose a new intrusion detection method that uses image conversion from network data flow to produce an RGB image that can be classified using advanced deep learning models. In this method, we proposed to use the decision tree algorithm to identify the important features, and a windowing and overlapping mechanism to convert the varying input size to a standard size image for the classifier. We then use a Vision Transfomer (ViT) classifier to classify the resulting image. Our experimental results show that we can achieve 98.5% accuracy in binary classification on the CIC IDS2017 dataset, and 96.3% on the UNSW-NB15 dataset, which is 8.09% higher than the next best algorithm, the Deep Belief Network with Improved Kernel-Based Extreme Learning (DBN-KELM) method. For multi-class classification, our proposed method can achieve a testing accuracy of 96.4%, which is 5.6% higher than the next best method, the DBN-KELM.


  • Structure-Aware Reinforcement Learning for Node-Overload Protection in Mobile Edge Computing

    by Anirudha Jitani, Aditya Mahajan, Zhongwen Zhu, Hatem Abou-Zeid, Emmanuel Thepie Fapi, Hakimeh Purmehdi

    Mobile Edge Computing (MEC) involves placing computational capability and applications at the edge of the network, providing benefits such as reduced latency, reduced network congestion, and improved performance of applications. The performance and reliability of MEC degrades significantly when the edge server(s) in the cluster are overloaded. In this work, an adaptive admission control policy to prevent edge node from getting overloaded is presented. This approach is based on a recently-proposed low complexity RL (Reinforcement Learning) algorithm called SALMUT (Structure-Aware Learning for Multiple Thresholds), which exploits the structure of the optimal admission control policy in multi-class queues for an average-cost setting. We extend the framework to work for node overload-protection problem in a discounted-cost setting. The proposed solution is validated using several scenarios mimicking real-world deployments in two different settings — computer simulations and a docker testbed. Our empirical evaluations show that the total discounted cost incurred by SALMUT is similar to state-of-the-art deep RL algorithms such as PPO (Proximal Policy Optimization) and A2C (Advantage Actor Critic) but requires an order of magnitude less time to train, outputs easily interpretable policy, and can be deployed in an online manner.


  • Joint Consensus Matrix Design and Resource Allocation for Decentralized Learning

    by Jingrong Wang, Ben Liang, Zhongwen Zhu, Emmanuel Thepie Fapi, Hardik Dalal

    In decentralized machine learning over a network of workers, each worker updates its local model as a weighted average of its local model and all models received from its neighbors. Efficient consensus weight matrix design and communication resource allocation can increase the training convergence rate and reduce the wall-clock training time. In this paper, we jointly consider these two factors and propose a novel algorithm termed Communication-Efficient Network Topology (CENT), which reduces the latency in each training iteration by removing unnecessary communication links. CENT preserves the training convergence rate while enforcing communication graph sparsity and avoiding selecting poor communication links. Numerical study with real-world machine learning data demonstrates the efficacy of the proposed solution and its performance advantage over state-of-the-art algorithms.


  • Multi-Objective GAN-Based Adversarial Attack Technique for Modulation Classifiers

    by Paulo Freitas de Araujo-Filho, Georges Kaddoum, Mohamed Naili, Emmanuel Thepie Fapi, Zhongwen Zhu

    Deep learning is increasingly being used for many tasks in wireless communications, such as modulation classification. However, it has been shown to be vulnerable to adversarial attacks, which introduce specially crafted imperceptible perturbations, inducing models to make mistakes. This letter proposes an input-agnostic adversarial attack technique that is based on generative adversarial networks (GANs) and multi-task loss. Our results show that our technique reduces the accuracy of a modulation classifier more than a jamming attack and other adversarial attack techniques. Furthermore, it generates adversarial samples at least 335 times faster than the other techniques evaluated, which raises serious concerns about using deep learning-based modulation classifiers.