Learning-Based Reconstruction Attacks on Coordinate-Obfuscated Point Clouds
本文评估了选择性坐标加密对基于机器学习的重建攻击的鲁棒性，通过PointNet和随机森林模型测试不同加密粒度下的安全性。
0 citationsRead paper
Selected work
Representative Papers
ExplainGuard: A Zero Trust Framework for Post-Hoc Explanation Integrity Guarantees in Blackbox XAI Models
为解决黑盒XAI模型后验解释完整性问题，提出ExplainGuard框架，采用零信任架构确保解释生成过程的验证与可信。
0 citationsRead paper
Malformer: A Multi-Modal Malware Detector Using Transformers
本文提出Malformer，一种结合文本、图像、图和音频四种模态的Windows恶意软件检测模型，通过多模态Transformer融合提升检测性能，解决了单一模态检测新威胁能力不足的问题。
0 citationsRead paper
Experimental Protocol Fingerprinting in Quantum Networks via Physical Layer Side Channel Analysis
This study addresses the unexplored vulnerability of quantum communication protocols to physical-layer side-channel attacks, specifically their distinguishability and associated security risks. The authors propose a non-invasive, passive side-channel analysis method that preserves quantum entanglement while extracting protocol-specific fingerprints from features such as single-photon detection statistics and optical power in photonic signals. By integrating machine learning for classification, the approach achieves 96% protocol identification accuracy under a 30:70 sampling ratio and maintains 70–89% accuracy even at a challenging 10:90 ratio. Crucially, Bell inequality tests confirm that entanglement remains intact throughout the process. This work provides the first experimental demonstration that protocol-level information can be leaked through non-intrusive side channels, thereby uncovering a novel threat vector to quantum communication security.
0 citationsRead paper
Experimental Side Channel Analysis of Protocol Stages in Quantum Identity Authentication
This study addresses a critical vulnerability in quantum identity authentication protocols, which are susceptible to side-channel attacks at the physical layer. By identifying the protocol phase, an adversary can bypass authentication and exfiltrate data. The work presents the first experimental side-channel analysis of such protocols, employing a custom-built quantum communication testbed to non-invasively capture photon arrival times and optical power via beam splitters. Leveraging feature engineering and machine learning models, the approach achieves high-accuracy protocol phase identification, attaining 98% accuracy (F1-score: 97%) at a 30% signal sampling rate and 96% accuracy (F1-score: 94%) at 10%. These results expose a novel class of security flaws and provide crucial empirical evidence for enhancing the physical-layer security design of quantum protocols.
0 citationsRead paper
Recent publications
Latest Papers
Learning-Based Reconstruction Attacks on Coordinate-Obfuscated Point Clouds
本文评估了选择性坐标加密对基于机器学习的重建攻击的鲁棒性，通过PointNet和随机森林模型测试不同加密粒度下的安全性。
0 citationsRead paper
ExplainGuard: A Zero Trust Framework for Post-Hoc Explanation Integrity Guarantees in Blackbox XAI Models
为解决黑盒XAI模型后验解释完整性问题，提出ExplainGuard框架，采用零信任架构确保解释生成过程的验证与可信。
0 citationsRead paper
Malformer: A Multi-Modal Malware Detector Using Transformers
本文提出Malformer，一种结合文本、图像、图和音频四种模态的Windows恶意软件检测模型，通过多模态Transformer融合提升检测性能，解决了单一模态检测新威胁能力不足的问题。
0 citationsRead paper
Experimental Protocol Fingerprinting in Quantum Networks via Physical Layer Side Channel Analysis
This study addresses the unexplored vulnerability of quantum communication protocols to physical-layer side-channel attacks, specifically their distinguishability and associated security risks. The authors propose a non-invasive, passive side-channel analysis method that preserves quantum entanglement while extracting protocol-specific fingerprints from features such as single-photon detection statistics and optical power in photonic signals. By integrating machine learning for classification, the approach achieves 96% protocol identification accuracy under a 30:70 sampling ratio and maintains 70–89% accuracy even at a challenging 10:90 ratio. Crucially, Bell inequality tests confirm that entanglement remains intact throughout the process. This work provides the first experimental demonstration that protocol-level information can be leaked through non-intrusive side channels, thereby uncovering a novel threat vector to quantum communication security.
0 citationsRead paper
Experimental Side Channel Analysis of Protocol Stages in Quantum Identity Authentication
This study addresses a critical vulnerability in quantum identity authentication protocols, which are susceptible to side-channel attacks at the physical layer. By identifying the protocol phase, an adversary can bypass authentication and exfiltrate data. The work presents the first experimental side-channel analysis of such protocols, employing a custom-built quantum communication testbed to non-invasively capture photon arrival times and optical power via beam splitters. Leveraging feature engineering and machine learning models, the approach achieves high-accuracy protocol phase identification, attaining 98% accuracy (F1-score: 97%) at a 30% signal sampling rate and 96% accuracy (F1-score: 94%) at 10%. These results expose a novel class of security flaws and provide crucial empirical evidence for enhancing the physical-layer security design of quantum protocols.
0 citationsRead paper