Nii-MALA: A Fast Metropolis-Adjusted Langevin Sampler with Radial Velocity Benchmarks
该研究提出Nii-MALA算法，利用高效并行化和自动微分技术提高高维分布采样效率，并通过天文数据验证其有效性。
0 citationsRead paper
Selected work
Representative Papers
Rethinking Battery-free Sensing Communication via Wake-up Radios
该研究通过结合微安唤醒无线电与低功耗实时时钟，解决了无电池传感器通信窗口短且不稳定的问题，增强了首次接触机会及后续通信的稳定性。
0 citationsRead paper
LH2Face: Loss function for Hard High-quality Face
Existing cosine-similarity- and Softmax-based face recognition methods exhibit insufficient discriminative power on challenging high-quality samples. To address this, we propose LH2Face, a novel loss function. Our approach models face features on the hypersphere using the von Mises–Fisher (vMF) distribution—replacing conventional Euclidean or cosine metrics—and introduces an uncertainty-aware adaptive margin function that jointly characterizes sample difficulty and quality. Furthermore, LH2Face integrates proxy-based classification loss with a face reconstruction task within a multi-task optimization framework. Evaluated on the IJB-B benchmark, LH2Face achieves 49.39% true positive rate at a false acceptance rate of 1e−4, outperforming the second-best method by 2.37%. This demonstrates substantial improvement in recognizing high-quality yet difficult samples.
0 citationsRead paper
Security Analysis of Thumbnail-Preserving Image Encryption and a New Framework
This paper identifies a fundamental security flaw in Thumbnail-Preserving Encryption (TPE): inherent multi-image thumbnail collisions that undermine cloud-based image search. To address this, we propose Multi-Factor Thumbnail-Preserving Encryption (MFTPE), a novel framework that (1) establishes the first theoretical model quantifying thumbnail collision probability—from block-level to *N*-image-level; (2) introduces a customizable feature construction method integrating sum, range, weighted mean, and geometric mean; and (3) rigorously validates reduced collision probability via theoretical security analysis and robustness experiments. MFTPE preserves thumbnail visual utility and search efficiency while effectively resisting facial detection and diverse noise attacks. It thus achieves synergistic enhancement of privacy protection and searchable functionality.
0 citationsRead paper
Recent publications
Latest Papers
Nii-MALA: A Fast Metropolis-Adjusted Langevin Sampler with Radial Velocity Benchmarks
该研究提出Nii-MALA算法，利用高效并行化和自动微分技术提高高维分布采样效率，并通过天文数据验证其有效性。
0 citationsRead paper
Rethinking Battery-free Sensing Communication via Wake-up Radios
该研究通过结合微安唤醒无线电与低功耗实时时钟，解决了无电池传感器通信窗口短且不稳定的问题，增强了首次接触机会及后续通信的稳定性。
0 citationsRead paper
LH2Face: Loss function for Hard High-quality Face
Existing cosine-similarity- and Softmax-based face recognition methods exhibit insufficient discriminative power on challenging high-quality samples. To address this, we propose LH2Face, a novel loss function. Our approach models face features on the hypersphere using the von Mises–Fisher (vMF) distribution—replacing conventional Euclidean or cosine metrics—and introduces an uncertainty-aware adaptive margin function that jointly characterizes sample difficulty and quality. Furthermore, LH2Face integrates proxy-based classification loss with a face reconstruction task within a multi-task optimization framework. Evaluated on the IJB-B benchmark, LH2Face achieves 49.39% true positive rate at a false acceptance rate of 1e−4, outperforming the second-best method by 2.37%. This demonstrates substantial improvement in recognizing high-quality yet difficult samples.
0 citationsRead paper
Security Analysis of Thumbnail-Preserving Image Encryption and a New Framework
This paper identifies a fundamental security flaw in Thumbnail-Preserving Encryption (TPE): inherent multi-image thumbnail collisions that undermine cloud-based image search. To address this, we propose Multi-Factor Thumbnail-Preserving Encryption (MFTPE), a novel framework that (1) establishes the first theoretical model quantifying thumbnail collision probability—from block-level to *N*-image-level; (2) introduces a customizable feature construction method integrating sum, range, weighted mean, and geometric mean; and (3) rigorously validates reduced collision probability via theoretical security analysis and robustness experiments. MFTPE preserves thumbnail visual utility and search efficiency while effectively resisting facial detection and diverse noise attacks. It thus achieves synergistic enhancement of privacy protection and searchable functionality.
0 citationsRead paper