Calmables: Demonstrating Closed-Loop Infrared Earables for Thermal Biofeedback and Relaxation Support
研究通过智能戒指监测心率，并使用闭环红外耳戴设备提供局部温暖提示，以支持生物反馈和放松，帮助用户在急性激活后恢复。
0 citationsRead paper
Selected work
Representative Papers
Face Re-morphing: Differential Morphing Attack Detection via Feature-Space Similarity Changes
This study addresses the performance limitations of existing face morphing attack detection methods that rely predominantly on static features. We propose a novel detection framework leveraging similarity variations induced by re-morphing operations. Specifically, this work pioneers the utilization of feature space response discrepancies caused by secondary morphing as complementary discriminative cues, achieving precise identification through image re-morphing generation and cosine similarity analysis. Extensive experiments demonstrate that the proposed method exhibits robust generalization across multiple datasets and models, with particularly superior performance on the AMSL and FEI Morph datasets. By effectively overcoming the bottlenecks inherent in traditional static detection approaches, this research establishes a new paradigm for defending against face morphing attacks.
0 citationsRead paper
XSA-MAD: Cross-modal Semantic Alignment for Morphing Attack Detection
This study addresses the limited generalizability of existing image manipulation detection methods by proposing a CLIP-based multimodal framework. The approach innovatively decouples manipulation into four interpretable attributes—identity, geometry, texture, and consistency—and constructs a unified semantic representation through progressive cross-modal alignment, enabling generation-agnostic, concept-level discrepancy modeling. By effectively capturing generation-invariant features, the method demonstrates strong generalization on the MAD22 and MorDIFF datasets, achieving an Equal Error Rate (EER) as low as 2.92% for GAN manipulations. Furthermore, it consistently outperforms state-of-the-art approaches in high-fidelity attack scenarios, significantly enhancing detection robustness against diverse and evolving synthetic media threats.
0 citationsRead paper
Cross-country structure of a sexual contact network derived from a commercial-sex review platform
Constrained by technological, ethical, and privacy limitations, large-scale sexual contact networks remain largely unobservable, impeding understanding of the cross-regional transmission dynamics of sexually transmitted infections. This study leverages over 1.43 million client–sex worker reviews from The Erotic Review platform (1999–2024) to construct, for the first time, a multinational bipartite network. Applying complex network methodologies—including connected component identification, small-world analysis, and degree correlation—we uncover its macroscopic connectivity structure. The network exhibits small-world properties and contains a giant connected component. A small number of high-frequency cross-border clients act as spatial bridges, while long-term users serve as temporal bridges, both significantly enhancing network cohesion. Interactions involving transgender female sex workers are highly concentrated among these highly active cross-border clients, revealing pronounced structural centralization.
0 citationsRead paper
Group invariance of $f$-divergences and the Fisher--Rao distance
This study investigates the invariance properties of distributional divergence measures within group-symmetric statistical models. By endowing both the sample and parameter spaces with a group action and assuming that density functions transform according to a multiplier representation, the authors integrate tools from group representation theory, transformation models, $f$-divergence analysis, and Fisher–Rao information geometry. They establish that all $f$-divergences and the Fisher–Rao distance are invariant under the induced group action. The key contribution lies in showing that such invariant divergences reduce to functions depending solely on the maximal invariants of the parameter pair. This framework is successfully extended to multivariate location-scale families, where the invariant geometric structure of the parameter space is characterized via double coset decompositions.
0 citationsRead paper
Recent publications
Latest Papers
Calmables: Demonstrating Closed-Loop Infrared Earables for Thermal Biofeedback and Relaxation Support
研究通过智能戒指监测心率，并使用闭环红外耳戴设备提供局部温暖提示，以支持生物反馈和放松，帮助用户在急性激活后恢复。
0 citationsRead paper
Face Re-morphing: Differential Morphing Attack Detection via Feature-Space Similarity Changes
This study addresses the performance limitations of existing face morphing attack detection methods that rely predominantly on static features. We propose a novel detection framework leveraging similarity variations induced by re-morphing operations. Specifically, this work pioneers the utilization of feature space response discrepancies caused by secondary morphing as complementary discriminative cues, achieving precise identification through image re-morphing generation and cosine similarity analysis. Extensive experiments demonstrate that the proposed method exhibits robust generalization across multiple datasets and models, with particularly superior performance on the AMSL and FEI Morph datasets. By effectively overcoming the bottlenecks inherent in traditional static detection approaches, this research establishes a new paradigm for defending against face morphing attacks.
0 citationsRead paper
XSA-MAD: Cross-modal Semantic Alignment for Morphing Attack Detection
This study addresses the limited generalizability of existing image manipulation detection methods by proposing a CLIP-based multimodal framework. The approach innovatively decouples manipulation into four interpretable attributes—identity, geometry, texture, and consistency—and constructs a unified semantic representation through progressive cross-modal alignment, enabling generation-agnostic, concept-level discrepancy modeling. By effectively capturing generation-invariant features, the method demonstrates strong generalization on the MAD22 and MorDIFF datasets, achieving an Equal Error Rate (EER) as low as 2.92% for GAN manipulations. Furthermore, it consistently outperforms state-of-the-art approaches in high-fidelity attack scenarios, significantly enhancing detection robustness against diverse and evolving synthetic media threats.
0 citationsRead paper
Cross-country structure of a sexual contact network derived from a commercial-sex review platform
Constrained by technological, ethical, and privacy limitations, large-scale sexual contact networks remain largely unobservable, impeding understanding of the cross-regional transmission dynamics of sexually transmitted infections. This study leverages over 1.43 million client–sex worker reviews from The Erotic Review platform (1999–2024) to construct, for the first time, a multinational bipartite network. Applying complex network methodologies—including connected component identification, small-world analysis, and degree correlation—we uncover its macroscopic connectivity structure. The network exhibits small-world properties and contains a giant connected component. A small number of high-frequency cross-border clients act as spatial bridges, while long-term users serve as temporal bridges, both significantly enhancing network cohesion. Interactions involving transgender female sex workers are highly concentrated among these highly active cross-border clients, revealing pronounced structural centralization.
0 citationsRead paper
Group invariance of $f$-divergences and the Fisher--Rao distance
This study investigates the invariance properties of distributional divergence measures within group-symmetric statistical models. By endowing both the sample and parameter spaces with a group action and assuming that density functions transform according to a multiplier representation, the authors integrate tools from group representation theory, transformation models, $f$-divergence analysis, and Fisher–Rao information geometry. They establish that all $f$-divergences and the Fisher–Rao distance are invariant under the induced group action. The key contribution lies in showing that such invariant divergences reduce to functions depending solely on the maximal invariants of the parameter pair. This framework is successfully extended to multivariate location-scale families, where the invariant geometric structure of the parameter space is characterized via double coset decompositions.
0 citationsRead paper