AspisAI: A Canonical, Machine-Interpretable Governance Framework for Automated Multi-Standard Compliance Monitoring
本文提出AspisAI框架，通过将多种标准要求转化为机器可解释的模型并评估证据，解决多标准合规监测成本高、一致性差的问题。
0 citationsRead paper
Selected work
Representative Papers
Evidence-Aware Protein Complex Detection: Methods, Benchmarks, and Reproducibility Challenges
Accurately identifying protein complexes from noisy, incomplete, and context-dependent protein–protein interaction (PPI) networks remains challenging, compounded by a lack of standardized and reproducible evaluation practices. This work systematically reviews and evaluates methods that integrate PPI network topology with multi-source biological evidence—such as Gene Ontology (GO) annotations, gene expression profiles, and subcellular localization—and proposes transparent, evidence-aware graph models as the current best trade-off. It is the first to systematically highlight the critical roles of GO circularity, overlap-aware metrics, and uncertainty quantification in complex detection evaluation. The study advocates for unified benchmarks, explicit circularity control, executable software packages, and rigorous benchmarking of advanced architectures—including deep learning, hypergraph, and dynamic heterogeneous models—to steer the field toward reproducible and biologically plausible methodologies.
0 citationsRead paper
One Transit Is All You Need: Detecting Exoplanets Through Learned Stellar Behaviour with EXOVEIL
This study addresses the challenge of detecting single-epoch exoplanet transits, which are inherently non-periodic and thus evade conventional transit detection methods reliant on periodic signals. The authors propose a self-supervised world model based on a Transformer architecture that learns the intrinsic stellar light-curve behavior through masked prediction. By identifying anomalous transit-like signals via prediction residuals—without requiring phase folding—the method enables effective detection of singular transit events. This approach achieves, for the first time, reliable identification of single transits, supports zero-shot transfer across tasks, and incorporates conformal prediction to enhance result reliability. Evaluated on Kepler data, the model attains an AUC of 0.938, recovers 32% of injected single transits with a depth of 1000 ppm, uncovers 179 new blind-search candidates, and demonstrates 100% zero-shot recovery of known TESS planets.
0 citationsRead paper
ECHO-PPI: Trustworthy AI for Evidence-Bundled Detection of Overlapping Protein Modules in Protein-Protein Interaction Networks
This study addresses the limited interpretability of existing methods for detecting functional modules in protein–protein interaction (PPI) networks, which often fail to explain why a protein is assigned to a module or to distinguish core, peripheral, and ambiguous members. To overcome these limitations, the authors propose ECHO-PPI, a novel framework that introduces an evidence-bundling mechanism and hierarchical confidence labels. By integrating weighted network topology, semantic protein profiles, and Gene Ontology evidence, ECHO-PPI enables interpretable and auditable detection of overlapping modules. The method maintains competitive detection performance while substantially enhancing the reliability and explainability of predictions, thereby facilitating downstream biological validation and manual prioritization.
0 citationsRead paper
STRIDE-AI: A Threat Modeling Framework for Generative AI Security Assessment
Traditional cybersecurity approaches struggle to address the probabilistic nature of generative AI, rendering systems vulnerable to emerging threats such as model inversion, data poisoning, and prompt injection attacks. This work proposes the first STRIDE-based threat modeling framework tailored specifically for generative AI, which seamlessly integrates the NIST AI Risk Management Framework (RMF) with the OWASP LLM Top 10 vulnerability taxonomy. The framework establishes a comprehensive six-phase AI security assessment lifecycle and is accompanied by a dedicated web-based tool. Evaluated in a sandboxed LLM chatbot environment, the approach significantly reduces attack success rates from 80% to 15%, effectively bridging the gap between high-level risk governance and low-level technical defenses.
0 citationsRead paper
Recent publications
Latest Papers
AspisAI: A Canonical, Machine-Interpretable Governance Framework for Automated Multi-Standard Compliance Monitoring
本文提出AspisAI框架，通过将多种标准要求转化为机器可解释的模型并评估证据，解决多标准合规监测成本高、一致性差的问题。
0 citationsRead paper
Evidence-Aware Protein Complex Detection: Methods, Benchmarks, and Reproducibility Challenges
Accurately identifying protein complexes from noisy, incomplete, and context-dependent protein–protein interaction (PPI) networks remains challenging, compounded by a lack of standardized and reproducible evaluation practices. This work systematically reviews and evaluates methods that integrate PPI network topology with multi-source biological evidence—such as Gene Ontology (GO) annotations, gene expression profiles, and subcellular localization—and proposes transparent, evidence-aware graph models as the current best trade-off. It is the first to systematically highlight the critical roles of GO circularity, overlap-aware metrics, and uncertainty quantification in complex detection evaluation. The study advocates for unified benchmarks, explicit circularity control, executable software packages, and rigorous benchmarking of advanced architectures—including deep learning, hypergraph, and dynamic heterogeneous models—to steer the field toward reproducible and biologically plausible methodologies.
0 citationsRead paper
One Transit Is All You Need: Detecting Exoplanets Through Learned Stellar Behaviour with EXOVEIL
This study addresses the challenge of detecting single-epoch exoplanet transits, which are inherently non-periodic and thus evade conventional transit detection methods reliant on periodic signals. The authors propose a self-supervised world model based on a Transformer architecture that learns the intrinsic stellar light-curve behavior through masked prediction. By identifying anomalous transit-like signals via prediction residuals—without requiring phase folding—the method enables effective detection of singular transit events. This approach achieves, for the first time, reliable identification of single transits, supports zero-shot transfer across tasks, and incorporates conformal prediction to enhance result reliability. Evaluated on Kepler data, the model attains an AUC of 0.938, recovers 32% of injected single transits with a depth of 1000 ppm, uncovers 179 new blind-search candidates, and demonstrates 100% zero-shot recovery of known TESS planets.
0 citationsRead paper
ECHO-PPI: Trustworthy AI for Evidence-Bundled Detection of Overlapping Protein Modules in Protein-Protein Interaction Networks
This study addresses the limited interpretability of existing methods for detecting functional modules in protein–protein interaction (PPI) networks, which often fail to explain why a protein is assigned to a module or to distinguish core, peripheral, and ambiguous members. To overcome these limitations, the authors propose ECHO-PPI, a novel framework that introduces an evidence-bundling mechanism and hierarchical confidence labels. By integrating weighted network topology, semantic protein profiles, and Gene Ontology evidence, ECHO-PPI enables interpretable and auditable detection of overlapping modules. The method maintains competitive detection performance while substantially enhancing the reliability and explainability of predictions, thereby facilitating downstream biological validation and manual prioritization.
0 citationsRead paper
STRIDE-AI: A Threat Modeling Framework for Generative AI Security Assessment
Traditional cybersecurity approaches struggle to address the probabilistic nature of generative AI, rendering systems vulnerable to emerging threats such as model inversion, data poisoning, and prompt injection attacks. This work proposes the first STRIDE-based threat modeling framework tailored specifically for generative AI, which seamlessly integrates the NIST AI Risk Management Framework (RMF) with the OWASP LLM Top 10 vulnerability taxonomy. The framework establishes a comprehensive six-phase AI security assessment lifecycle and is accompanied by a dedicated web-based tool. Evaluated in a sandboxed LLM chatbot environment, the approach significantly reduces attack success rates from 80% to 15%, effectively bridging the gap between high-level risk governance and low-level technical defenses.
0 citationsRead paper