Selected work

Representative Papers

PriEval-Protect: A Unified Framework for Privacy Evaluation and Protection in Healthcare Systems

Jul 15, 2026

This work addresses the limitations of current healthcare data privacy compliance approaches, which rely heavily on manual processes and treat policy auditing and technical risk assessment in isolation, leading to inefficiency and error-proneness. To overcome these challenges, the authors propose PriEval-Protect, a two-stage framework that integrates legal large language models with data-level privacy metrics during the evaluation phase. By leveraging retrieval-augmented generation (RAG), cryptographic identification, and analytic hierarchy process (AHP) weighting, the framework produces an interpretable, composite risk score. In the protection phase, it dynamically recommends mitigation strategies—such as federated learning or differential privacy—based on this score. Validated on real-world hospital data, PriEval-Protect enables synergistic compliance with GDPR and HIPAA, delivers precise risk assessment, and offers explainable privacy safeguards, significantly enhancing the automation and consistency of privacy governance.

0 citationsRead paper

STCALIR: Semi-Synthetic Test Collection for Algerian Legal Information Retrieval

Apr 01, 2026

This study addresses the challenges of legal information retrieval in Algeria, where high-quality corpora and relevance judgments are scarce, and manually constructing test collections is prohibitively expensive. Building upon the Cranfield paradigm, the authors propose a multi-stage automated retrieval and filtering approach combined with a semi-synthetic relevance judgment generation technique, which substantially reduces the need for human annotation. The method achieves a 99% reduction in annotation effort while preserving evaluation reliability, attaining a Hit@10 score of 0.785. System rankings demonstrate strong agreement with human assessments, as evidenced by Kendall’s τ of 0.89 and Spearman’s ρ of 0.92. This work offers a reproducible, low-cost framework for test collection construction in low-resource legal domains.

0 citationsRead paper

Neural Architecture Search with Mixed Bio-inspired Learning Rules

Jul 17, 2025

Biological neural networks offer advantages in robustness, energy efficiency, and physiological interpretability, yet bio-inspired models often lag behind backpropagation (BP)-based counterparts in accuracy and scalability. To address this, we propose an inter-layer heterogeneous biologically plausible learning rule hybridization mechanism—enabling adaptive selection of diverse neurobiological rules (e.g., STDP, Hippo) per layer—and introduce a dedicated neural architecture search (NAS) framework co-optimizing both network topology and rule assignment. This approach breaks the constraint of uniform learning rules across layers. Empirical evaluation demonstrates state-of-the-art performance among biologically inspired models on CIFAR-10, CIFAR-100, and ImageNet; notably, certain configurations surpass comparably sized BP models in accuracy while preserving inherent robustness and ultra-low energy consumption.

0 citationsRead paper

Recent publications

Latest Papers

PriEval-Protect: A Unified Framework for Privacy Evaluation and Protection in Healthcare Systems

Jul 15, 2026

This work addresses the limitations of current healthcare data privacy compliance approaches, which rely heavily on manual processes and treat policy auditing and technical risk assessment in isolation, leading to inefficiency and error-proneness. To overcome these challenges, the authors propose PriEval-Protect, a two-stage framework that integrates legal large language models with data-level privacy metrics during the evaluation phase. By leveraging retrieval-augmented generation (RAG), cryptographic identification, and analytic hierarchy process (AHP) weighting, the framework produces an interpretable, composite risk score. In the protection phase, it dynamically recommends mitigation strategies—such as federated learning or differential privacy—based on this score. Validated on real-world hospital data, PriEval-Protect enables synergistic compliance with GDPR and HIPAA, delivers precise risk assessment, and offers explainable privacy safeguards, significantly enhancing the automation and consistency of privacy governance.

0 citationsRead paper

STCALIR: Semi-Synthetic Test Collection for Algerian Legal Information Retrieval

Apr 01, 2026

This study addresses the challenges of legal information retrieval in Algeria, where high-quality corpora and relevance judgments are scarce, and manually constructing test collections is prohibitively expensive. Building upon the Cranfield paradigm, the authors propose a multi-stage automated retrieval and filtering approach combined with a semi-synthetic relevance judgment generation technique, which substantially reduces the need for human annotation. The method achieves a 99% reduction in annotation effort while preserving evaluation reliability, attaining a Hit@10 score of 0.785. System rankings demonstrate strong agreement with human assessments, as evidenced by Kendall’s τ of 0.89 and Spearman’s ρ of 0.92. This work offers a reproducible, low-cost framework for test collection construction in low-resource legal domains.

0 citationsRead paper

Neural Architecture Search with Mixed Bio-inspired Learning Rules

Jul 17, 2025

Biological neural networks offer advantages in robustness, energy efficiency, and physiological interpretability, yet bio-inspired models often lag behind backpropagation (BP)-based counterparts in accuracy and scalability. To address this, we propose an inter-layer heterogeneous biologically plausible learning rule hybridization mechanism—enabling adaptive selection of diverse neurobiological rules (e.g., STDP, Hippo) per layer—and introduce a dedicated neural architecture search (NAS) framework co-optimizing both network topology and rule assignment. This approach breaks the constraint of uniform learning rules across layers. Empirical evaluation demonstrates state-of-the-art performance among biologically inspired models on CIFAR-10, CIFAR-100, and ImageNet; notably, certain configurations surpass comparably sized BP models in accuracy while preserving inherent robustness and ultra-low energy consumption.

0 citationsRead paper