pii detection

This study addresses the central challenge in privacy-preserving classification: how to acquire useful individual information while avoiding the disclosure of sensitive attributes. For the first time, we formally integrate the privacy objective into the classification decision process, proving the resulting problem to be NP-complete. To tackle this, we propose an interactive querying framework based on generalized decision trees that optimizes question strategies to reduce uncertainty about the target attribute while constraining the leakage of private information. We develop an efficient heuristic solver combining greedy and genetic algorithms to navigate this trade-off. Experimental results demonstrate that our approach achieves an optimal balance between utility and privacy within acceptable computational time, offering a practical solution for non-intrusive personnel screening and customer data collection scenarios.

Privacy policies are often lengthy and ambiguous, impeding users’ comprehension of data collection, sharing, and tracking practices—thereby undermining privacy transparency and informed decision-making. To address this, we propose a novel analytical framework integrating graph neural networks, semantic parsing, and interactive visualization. First, we construct a fine-grained semantic graph model from policy text using natural language processing. Second, we apply graph mining alongside t-SNE/PCA for topic clustering and interpretable dimensionality reduction. Finally, an interactive knowledge graph interface enables risk identification and compliance auditing. Our key contribution is the first synergistic application of dynamic graph visualization and graph mining to structured privacy policy analysis. This approach significantly improves textual readability (42% gain in structural clarity) and pattern recognition capability, uncovering cross-platform data-sharing and covert tracking commonalities. The framework provides scalable technical support for regulatory oversight and user empowerment.

Existing privacy research is fragmented across isolated technical domains (e.g., CV, NLP, networking), failing to address real-world, cross-contextual privacy concerns. Method: We propose a human-centered privacy modeling framework grounded in Contextual Integrity (CI) theory—first systematically integrating CI into large language models (LLMs). Our approach constructs the first multi-ontology privacy checklist incorporating social identities, sensitive attributes, and the full HIPAA regulatory framework. Leveraging expert-annotated, multi-source ontology fusion, it extends beyond traditional PII definitions to enable context-aware, human-interpretable privacy assessment. Contribution/Results: Experiments demonstrate LLMs’ efficacy in structured regulatory comprehension and context-sensitive privacy reasoning. Our framework establishes a novel paradigm for generalizable, cross-domain privacy risk identification—bridging theoretical privacy principles with scalable, deployable AI-driven assessment.

Modern encryption protocols (e.g., TLS 1.3) have rendered traditional network traffic classification (NTC) ineffective, while existing ML-based NTC studies suffer from outdated datasets, flawed assumptions, and methodological biases. Method: We conduct a systematic mapping study (SoK), establishing a taxonomy of NTC design choices and benchmarking practices; we perform the first large-scale empirical validation—348 feature-masking ablation experiments—to rigorously test foundational assumptions, exposing critical flaws such as training on plaintext traffic and overfitting to obsolete datasets. Contribution/Results: We demonstrate that state-of-the-art classifiers consistently fail on real-world encrypted traffic; we propose a reproducible best-practice guideline for NTC evaluation; and we delineate three key future directions: (1) focusing on realistic encrypted scenarios, (2) enabling dynamic dataset curation and updating, and (3) adopting causality-aware, domain-informed feature engineering.

This study addresses the significant gap between academic research and real-world deployment in network intrusion detection systems (NIDS), which stems from a lack of consensus on the fundamental characteristics of NIDS and consequently leads to inconsistent evaluation benchmarks. Through a systematic survey (SoK), this work formally defines the intrinsic properties of NIDS, critically examines prevailing evaluation methodologies, and employs reproducible case studies to expose the disconnect between theoretical research and operational practice. Building on these insights, the paper proposes foundational principles and concrete recommendations for reframing NIDS research through the lens of security operations, aiming to align academic inquiry more closely with real-world requirements and provide actionable methodological guidance for future work.

Clinical electronic health records (EHRs) contain critical anomalous events that evade detection by rule-based systems—particularly subtle, “apparently normal” events with substantial prognostic impact. Method: We propose an information-theoretic, foundation-model–driven anomaly detection framework that quantifies token- and event-level informativeness via context-aware entropy estimation and event saliency scoring. Contribution/Results: This work pioneers the integration of information theory into clinical event detection, overcoming reliance on handcrafted rules. By coupling informativeness scoring with interpretable attribution techniques, it provides event-level explanations for prognostic models. Experiments demonstrate that high-informativeness events identified by our method significantly improve downstream outcome prediction (average AUC gain of 0.08), while low-informativeness events can be safely pruned without performance degradation—validating the clinical utility and decision-support value of informativeness as a biomarker.

Individuals and organizations lack systematic awareness of interdependent privacy risks, hindering proactive mitigation of cascading data breaches. Method: Leveraging over 5,000 real-world identity theft cases, we construct the first empirically grounded identity attribute ecosystem graph model, explicitly representing disclosure dependencies among personally identifiable information (PII) as a directed graph; we then propose a graph neural network (GNN)-based privacy risk prediction framework to quantitatively assess cascading leakage pathways—i.e., from single-attribute exposure to multi-attribute compromise. Contribution/Results: Experiments demonstrate that our approach significantly outperforms conventional feature-engineering models in both predictive accuracy and interpretability. It provides a theoretically grounded, practically deployable tool for pre-incident privacy risk forecasting, enabling actionable early warnings and supporting privacy-by-design strategies.

This work addresses the precision-recall trade-off in network intrusion detection, arising from the inherent diversity of cyber threats. Methodologically, we propose a novel detection framework that deeply integrates a cybersecurity incident taxonomy into the architectural design of detection networks. Guided by the taxonomy’s hierarchical semantic structure, we jointly leverage ontology-based analysis and controlled simulations to systematically identify the optimal operational equilibrium for detection strategies. Our key contribution is the first principled elevation of taxonomies from static labeling tools to structural priors embedded within detection models—explicitly encoding semantic relationships and evolutionary pathways among threat behaviors. Empirical evaluation across multiple public benchmark datasets demonstrates substantial improvements in holistic detection performance (average F1-score gain of 12.7%). Moreover, our analysis uncovers fundamental theoretical limits on detection set construction and establishes an interpretable pathway for performance optimization.

Existing sensitive data detection methods predominantly focus on personally identifiable information (PII), overlooking the contextual dependency of data sensitivity. This paper proposes a novel context-aware paradigm for sensitive data detection, introducing— for the first time—type contextualization and domain contextualization mechanisms. The approach integrates semantic type identification, document-level contextual modeling, sensitive rule retrieval, and large language model (LLM)-driven reasoning into an end-to-end detection framework. Evaluated on non-standard data domains—such as humanitarian datasets—the method achieves a 94% recall in type contextualization, outperforming commercial tools by 31 percentage points. Domain contextualization significantly enhances adaptability to complex, real-world scenarios. Furthermore, LLM-generated interpretive explanations substantially improve inter-annotator agreement during manual review. Collectively, this work advances sensitive data detection by grounding sensitivity assessment in rich, multi-granular contextual signals rather than static, syntax-driven patterns.

This study identifies critical flaws in current post-hoc privacy filters applied to synthetic chest X-ray data: they exhibit high sensitivity to real images but suffer from low specificity and poor consistency, failing reliably to detect near-duplicate samples generated from training data—thereby introducing false security guarantees. To address this, we propose the first tripartite evaluation framework for synthetic medical imaging, systematically assessing filters along three dimensions—sensitivity, specificity, and consistency—on both real and synthetic images. Empirical evaluation demonstrates that existing methods cannot robustly prevent training-data leakage and fall short of clinical-grade privacy assurance requirements. Our work not only quantifies key technical bottlenecks in privacy filtering but also establishes a new benchmark for privacy assessment of synthetic medical data, providing both theoretical foundations and practical standards for designing next-generation, robust, and verifiable healthcare privacy-preserving technologies.