Guoming Zhang

RadioShock: Over-the-Air Adversarial Attacks on Wireless Communication

Wenhao Li, Chenxu Li, Guoming Zhang, Zhijian Huang et autres

There is an emerging trend of using deep learning (DL) to handle complex tasks in wireless communication systems. However, recent research suggests that DL-enabled communication systems are vulnerable to adversarial attacks. Fortunately, most of these attacks are simulation-based, incapable of handling realistic …

cn, us, sg (code pays fourni par la source)

Special Issue on Securing the Internet of Things: Challenges, Innovations, and Practical Solutions

Yue Zhang, Xiaoshuang Xing, Guoming Zhang, Rajiv Kumar Ranjan

cn, gb (code pays fourni par la source)

Vehisper: Eavesdropping on In-Vehicle Audio via Secondary Magnetic Leakage

Heqiang Fu, Huiting Zhang, Guoming Zhang, Yanni Yang et autres

Modern vehicles are widely equipped with in-vehicle audio systems, and drivers routinely play music, podcasts, navigation prompts, phone calls, and various voice services during daily commutes. Due to the enclosed nature of the vehicle cabin and the presence of substantial ambient noise, …

cn (code pays fourni par la source)

Palm Vein Reconstruction From Electromagnetic Side-Channel Emissions

Zihang Lu, Xiang Gao, Ning Gao, Guoming Zhang et autres

Palm vein recognition has gained traction in secure authentication due to its unique, stable, and inherently concealed biometric characteristics. However, the electromagnetic (EM) emissions from subcutaneous vein imaging sensors (SVIS) may unintentionally leak sensitive biometric information, fundamentally challenging the assumed security guarantees. …

cn (code pays fourni par la source)

RFInv: Uncovering Sensitive Data in RF Sensing Systems via Model Inversion

Mingda Han, Huanqi Yang, Yuting Yang, Guoming Zhang et autres

Deep learning has significantly advanced Radio Frequency (RF) sensing, leading to extensive research and practical applications in both academia and industry. However, these advancements have also introduced potential privacy and security threats to RF sensing data. In this paper, we present RFInv, …

cn, hk (code pays fourni par la source)

Adversarial Attacks on Closed Box Speech Recognition Systems via Laser Injection

Hailong Zhang, Guoming Zhang, Xiaohui Ma, Zhijie Xiang et autres

Audio adversarial perturbations are designed to remain imperceptible to humans while deceiving automatic speech recognition (ASR) models. However, operating within the audible frequency range makes existing methods partially detectable in practice. In this paper, we present LaserAdv, a laser-based adversarial attack that …

cn (code pays fourni par la source)

THOR : Holistic Roadside Traffic Monitoring with a Single mmWave Radar

Wei Xu, Yanni Yang, Weijun Zou, Zezhao Li et autres

FDsign: A Signature Verification Method Based on Feature Fusion and Time-Series Diffusion Model

Weihao Zhang, Guoming Zhang, Xiaoyu Ji, Haiming Chen et autres

Handwritten signature verification has emerged as the predominant authentication method for paper-based transactions in critical sectors including financial services, legal documentation, and banking operations. While offline approaches utilizing static signature images remain vulnerable to forgery attacks, online methods demonstrate greater reliability by …

cn (code pays fourni par la source)

Microbial variations in the multibody sites of the pancreatic ductal adenocarcinoma patients

Yiqing Zhu, Mengfan Zhi, Xiao Liang, Lixiang Li et autres

OBJECTIVES: We aimed to investigate microbial characteristics across multibody sites of pancreatic ductal adenocarcinoma (PDAC) and identify novel microbial markers for PDAC. METHOD: We conducted 16S ribosomal RNA amplicon sequencing on saliva, duodenal fluid, and pancreatic tissue obtained via endoscopic ultrasound-guided fine …

cn (code pays fourni par la source)

Trigger-free and data-free backdoor attacks on deep neural networks

Jiahao Wang, Xianglong Zhang, Huanle Zhang, Xiaobo Ma et autres

近年来，随着深度神经网络广泛部署于众多关键领域，关于深度神经网络的后门攻击逐渐成为亟待关注的安全隐患。然而，多数后门攻击普遍假设攻击者能够获取目标模型的原始训练数据作为辅助数据集，这显著限制了攻击方法在现实场景中的实用性。此外，依赖显式触发器以激活后门的设计亦削弱了后门的隐蔽性。为此，本文提出一种无需训练数据与触发器的后门攻击方法，以提升后门攻击的实用性与隐蔽性。所提出方法基于一种新颖的微调框架，其将恶意数据的语义融入攻击者指定类别的特征表示中，从而实现无触发器恶意样本被稳定误分类至攻击目标类。同时，为保持模型在正常样本上的原始性能，本文引入知识蒸馏机制以替代传统训练数据，并结合弹性权重约束设计参数重要性评估机制，引导模型在保留原有知识的同时实现有效注入。我们在三个具有代表性的真实数据集上对所提方法进行了系统实证研究，实验结果验证了该攻击的有效性、隐蔽性与实用性。此外，本文还进一步探索了辅助数据集与模型反演技术在提升攻击能力方面的潜力。

Gaussian estimates for fundamental solutions of higher-order parabolic equations with time-independent coefficients

Guoming Zhang

We study the De Giorgi-Moser-Nash estimates of higher-order parabolic equations in divergence form with complex-valued, measurable, bounded, uniformly elliptic (in the sense of G$\mathring{a}$rding inequality) and time-independent coefficients. We also obtain Gaussian upper bounds and Hölder regularity estimates for the fundamental solutions …

Hidden pitfalls of walled-off necrosis: occult intracavitary artery in endoscopic transgastric necrosectomy

Miaomiao Ma, Xiaoya Wang, Guoming Zhang, Ruiguang Ma et autres

cn (code pays fourni par la source)