Search Results (66)

Background/Objectives: Discharge letters are essential for continuity of care but are often time-consuming to prepare and variable in quality. Large language models (LLMs) may help standardize and support this process, yet evidence in ophthalmology remains limited. This study compared the quality of resident-written and GPT-5.2-generated ophthalmology discharge letters derived from the same de-identified clinical data. Methods: This retrospective blinded study was conducted at a tertiary hospital in Croatia. For 146 consecutive inpatient discharges, original resident-written letters were paired with GPT-5.2-generated letters created using a standardized prompt; 142 complete pairs were available for the primary analysis. Three board-certified ophthalmologists evaluated anonymized letters using a structured assessment of accuracy, completeness, clarity/structure, tone/professional phrasing, conciseness, global quality, errors, omissions, and key content elements. Results: In the primary paired analysis, GPT-5.2-generated letters performed similarly to resident-written letters across accuracy, completeness, clarity/structure, errors, omissions, and overall quality. GPT-5.2-generated letters received higher ratings for tone/professional phrasing, whereas resident-written letters were rated as more concise, although inter-rater agreement was poor on these stylistic domains (at or below chance for conciseness) and these findings should therefore be interpreted as exploratory. Resident-written letters more often documented operations, while GPT-5.2-generated letters more consistently included findings. Reviewer-adjusted sensitivity analyses were less favorable to GPT-5.2 for several domains. Conclusions: GPT-5.2-generated ophthalmology discharge letters showed similar performance to resident-written letters in several evaluated domains in the primary paired analysis, but differences in specific content elements and less favorable sensitivity analyses indicate that clinician oversight remains necessary to ensure accuracy, procedural completeness, and clinical usability. Full article

Tor-based hidden services host substantial criminal infrastructure, yet de-anonymization research remains fragmented across heterogeneous techniques. No prior work has organized these techniques into a unified taxonomy oriented toward forensic investigation. This paper proposes a five-layer vulnerability taxonomy for Tor hidden services, distinguishing network-level (L1), application-level (L2), side-channel (L3), operational-security-failure (L4), and ecosystem-level (L5) categories. The taxonomy is derived from a structured review of literature published between 2002 and 2024. We further propose a Traceability Evaluation Framework (TEF) that scores 11 vulnerability types along three dimensions: Applicability, Technical Difficulty, and Legal Admissibility. The TEF dimension weights are derived through Analytic Hierarchy Process elicitation from a five-member expert panel of cybercrime investigators, digital forensics researchers, and a legal scholar. The resulting weights of (0.385, 0.204, 0.412) for Applicability, inverted Technical Difficulty, and Legal Admissibility prove robust to ±0.10 perturbations in sensitivity analysis. Under this framework, four application-layer (L2) and operational-security-failure (L4) vulnerabilities receive the highest traceability scores (TS ≥ 2.80), while two network-level (L1) attacks and one side-channel (L3) technique fall to the lowest tier. The framework integrates technical exploitability with legal admissibility constraints across U.S., EU, and other evidentiary regimes, providing a structured reference for investigators and a methodological foundation for case-based empirical validation in future work. Full article

The Model Context Protocol (MCP) has become the de facto standard for connecting large language models (LLMs) to external tools, and its remote deployment mode lets users add third-party servers with a single URL—shifting a substantial portion of the host’s attack surface to infrastructure operated by anonymous parties. Existing MCP security work has concentrated on tool-description poisoning and studied individual techniques in isolation, leaving it unclear what a malicious remote server can accomplish across its full surface. In this paper, we explore the malicious-server threat space along the axis of whether the host LLM participates in producing the harmful outcome, yielding two categories: LLM-passive attacks, which complete inside the server, and LLM-active attacks, which require the LLM to deliver the malicious content. We implement five scenarios spanning both categories—realizing each LLM-active scenario with both description-based and response-based variants against the same goal—and evaluate all configurations on ChatGPT, Claude Desktop, and Gemini CLI. We find that host-side filtering of MCP-bound data varies sharply across platforms (95% vs. 50% ASR on the same email request), that the description and response channels succeed on disjoint scenarios, and that successful attacks are almost never disclosed to the user. These findings suggest that defending remote MCP deployment requires a multi-layer approach combining host-side filtering, LLM-level response auditing, and user-visible output transparency. Full article

Background: The development of robust artificial intelligence (AI) models for diagnosing Temporomandibular Disorders (TMDs) is severely constrained by data scarcity and patient privacy regulations. Cone-beam computed tomography (CBCT), the gold standard for assessing osseous changes in the temporomandibular joint (TMJ), inherently contains sensitive biometric facial features, making de-identification difficult without losing critical anatomical information. This study aims to develop and evaluate TMJCTGenerator, a specialized latent diffusion model (LDM) framework designed to synthesize high-fidelity, diverse, and anonymous TMJ CBCT images. We hypothesize that this LDM approach can achieve superior anatomical fidelity and diversity compared to traditional generative adversarial network (GAN)- and variational autoencoder (VAE)-based methods, specifically in capturing fine osseous details within sagittal and coronal views of the mandibular condyle. Methods: A training dataset comprising 348 anonymized CBCT volumes was obtained in this retrospective comparative study to extract high-resolution sagittal and coronal regions of interest of the mandibular condyle. An independent test set of 39 anonymized CBCT volumes was further included. We developed a class-conditional LDM that integrates a pre-trained VAE for perceptual compression with a conditional U-Net for iterative denoising in the latent space. Performance was evaluated via qualitative anatomical fidelity assessment, Fréchet Inception Distance (FID), and a blinded Visual Turing test conducted by experienced clinicians to determine the distinguishability of synthetic images from real data. Results: Qualitative analysis revealed that TMJCTGenerator produced images with superior sharpness and anatomical consistency compared to baseline models, successfully reconstructing fine bone structures essential for diagnosing degenerative joint disease. TMJCTGenerator achieved lower FID scores than both VAE and GAN baselines. In the visual Turing test, clinicians were unable to reliably distinguish the generated images from real scans, and non-inferiority analysis confirmed that the synthetic data were statistically non-inferior to real data. Furthermore, TMJCTGenerator demonstrated the capability to generate diverse pathological conditions, ranging from normal anatomy to severe osteoarthritic changes. Conclusions: The proposed LDM framework effectively addresses the data scarcity and privacy bottlenecks in TMJ AI research by generating realistic, fully anonymous medical imaging data. TMJCTGenerator outperforms traditional generative methods in both visual fidelity and diversity, offering a viable solution for training downstream diagnostic algorithms. The source code and pre-trained models of TMJCTGenerator have been made open-source. Full article

Tor is an anonymization network that enables access to hidden services and protects user identity through layered encryption. While its core technology offers strong privacy, users can still be exposed through indirect attack methods or configuration mistakes. This research not only explores de-anonymization techniques but also provides a practical guide for constructing a fully functional experimental Tor environment using virtual machines. The custom-built testbed allows for safe simulation of attacks without impacting the public Tor network. Within this environment, three key information-gathering approaches were evaluated: (1) malware-based reverse shells that establish external communication, (2) malicious PDF and Office files used to trigger outbound connections, and (3) analysis of service misconfigurations that may reveal the IP address of hidden services. The results confirm that although the Tor network itself is resilient, user behavior, improper configurations, and insecure content handling can lead to significant privacy risks. By combining practical environment setup with real-world attack scenarios, this paper serves both as a reference for building experimental Tor networks and as a security-oriented analysis of known de-anonymization vectors. The findings emphasize the critical need for user awareness and precise configuration in privacy-focused technologies. Full article

This Data Descriptor presents an anonymized, shuffled dataset of creatinine-normalized urinary metabolite measurements from 73 Bulgarian children with autism spectrum disorder (ASD), released to support reuse in secondary analyses and cross-cohort comparisons. The public release represents a pathway-oriented 24-marker subset from a broader urinary diagnostic panel, assembled as a self-contained resource for investigators working in these metabolic domains. Spot urine results are provided as individual-level values after creatinine normalization; for trimethylamine, values below the limit of quantification (LOQ) were replaced with LOQ/2. The deposit contains measurements for 24 urinary markers grouped into three functional classes (neurotransmitters and aromatic amino acid precursors; one-carbon/methylation and vitamin-related metabolites; and energy metabolism/organic acids with microbiome-related amines). The underlying cohort comprised children aged 3–13 years, and no contemporaneous neurotypical control group was enrolled. Second-morning, midstream, acid-stabilized spot urine samples were collected within the provider’s workflow; metabolites were measured by LC–MS/MS, and spot urinary creatinine was measured enzymatically for normalization. The release includes the results table in both XLSX and CSV formats, a reference limits and units file for contextual interpretation, a data dictionary, a README, a changelog, and SHA-256 checksums for integrity verification. The public files contain de-identified analytical variables only and omit individual-level demographics, dates, standalone urinary creatinine, and richer clinical metadata to preserve anonymity. Full article

Profile images from social networks are a valuable source of data for AI analytics, but they contain biometric identifiers that pose serious privacy risks. The current face anonymization techniques often destroy semantic information, and generative de-identification methods are vulnerable to re-identification attacks. In this paper, we propose a template-driven multimodal face pseudonymization framework that allows for the privacy-preserving analysis of facial image data while retaining analytically relevant attributes. Our approach uses a FaceNet-based CelebA attribute classifier to extract fine-grained facial attributes and a DeepFace model to extract high-level demographic attributes. Rather than relying on stochastic large language models, we introduce deterministic template-based attribute-to-text conversion to ensure consistency and reproducibility and prevent unintended attribute hallucination. The resulting textual description serves as the sole conditioning input for Janus-Pro, a multimodal text-to-image generation model that synthesizes realistic yet non-identifiable face images. We evaluate our method on the CelebA dataset under a strong adversarial threat model, employing state-of-the-art face recognition systems to assess re-identification and linkability attacks. Our results demonstrate a substantial reduction in identity leakage while preserving semantic attributes. Full article

This article addresses a key evidence gap in urban safety policy in Colombia: the absence of publicly accessible microdata that jointly measure victimization, perception of security, and probability of sanctions among socioeconomically vulnerable residents. It aims to provide a clean, linkable dataset that enables analysis of variations in these issues across demographic and territorial groups in Cali (recently classified as the 29th most dangerous city worldwide, with 1028 and 1065 homicides in 2024 and 2025, respectively). It reports face-to-face survey data collected from 22 July to 16 August 2024, at Sistema de Identificación de Potenciales Beneficiarios de Programas Sociales (SISBEN) service points. The final dataset includes 2139 adults (aged 18–95 years) and combines (i) primary responses on perceived safety (e.g., public space safety and surveillance cameras), perceived likelihood of sanction, victimization, and self-protection measures with (ii) selected sociodemographic and household characteristics drawn from SISBEN IV records. Individual-level linkage was implemented using respondent identification at interviews, yielding an integrated anonymized file suitable for replication and secondary analysis. The dataset enables distributive analyses of insecurity (e.g., by sex, age, and ethnicity—including Afro-descendant populations) within a policy-relevant target group and supports evaluation and targeting of local interventions by providing individual-level indicators. Full article

The growth of the Internet of Things (IoT) has created many problems. A wise example is presented by the design of secure, efficient authentication and key agreement (AKA) protocols. A novel three-factor AKA protocol for the IoT is presented in this paper. The scheme integrates password, biometric, and device-based factors that achieved strong security, which gives anonymity to the user, achieves forward secrecy, and makes the scheme resilient to various attacks like replay, impersonation, and de-synchronization. It also adds a safe lost-password-reset functionality, which makes the protocol more usable. Security analysis proves its strength against the typical adversary, while performance evaluation shows that the solution is better than existing solutions in terms of computational and communication efficiency. The work proposes a practical and scalable security solution for IoT systems, which satisfies the high security standard but within the constraints of an IoT system. Full article

Background: Early evidence from a nursing home in Yepes (Toledo, Spain) indicated that antihistamines combined with azithromycin prevented deaths and hospitalizations during the first COVID-19 wave. Subsequent data from the Consorci Sanitari de Terrassa (CST) showed that patients chronically taking antihistamines had significantly reduced hospital admissions and mortality. However, a concerning rise in long COVID incidence (2–5%) after the third infection and a doubling of thrombosis rates in patients over 60 were observed. Objective: This study aimed to determine whether chronic antihistamine prescription is associated with a reduction in long COVID syndrome and thrombotic events. Methods: We analyzed anonymized data from the CST population (n = 192,651 as of March 2025). Variables included age, gender, chronic antihistamine use, number of chronic treatments (nT), COVID-19 vaccination status, SARS-CoV-2 infection history, long COVID (LC) incidence, and aggregated thrombotic events. Odds ratios (OR) were calculated using chi-square tests. Results: The prevalence of LC increased progressively with successive infections in the non-antihistamine group. No significant differences were found with the antihistamine group, which presented no LC cases among the 52 patients with three documented infections. Thrombotic events were significantly less frequent in antihistamine users with at least one chronic prescription (p < 0.0001). Conclusions: Results suggest a protective effect of antihistamines against thrombotic events. While confirmation via multicenter, randomized trials is needed, a pragmatic approach using antihistamines could be considered for symptomatic patients in the early stage of infection. Full article

This study presents the on-site public evaluation of Notre-Dame Whispers, a geolocated audio-guided tour that explores the sonic history of the Cathédrale Notre-Dame de Paris. The experience combines binaural reproduction, embodied storytelling, and historically informed soundscapes to immerse visitors in the cathedral’s past auditory environments. Drawing on virtually recreated acoustics, it reconstructs key components of Notre-Dame’s sound heritage, including the medieval construction site, early polyphonic chant, and the contemporary urban soundscape. An on-site evaluation was conducted to assess visitor engagement, usability, and the perceived authenticity of the reconstructed soundscapes. A mixed-methods approach integrated questionnaire responses, semi-structured interviews, and anonymized user analytics collected through the mobile application. Results indicate a high level of immersion, with participants particularly valuing the spatialised audio design and narrative depth. However, challenges were identified regarding GPS-based triggering reliability and the difficulty of situational interpretation in complex spatial environments. These findings offer insights into public reception of immersive heritage audio experiences and inform future developments in digital cultural mediation. Full article

Instant messaging (IM) applications are ubiquitous, and while end-to-end encryption protects message content, traffic metadata remains observable. This paper proposes a traffic correlation framework for IM call services under a passive ISP-level threat model to infer communication parties from encrypted traffic. The framework extracts and matches metadata from sustained, bidirectional call flows and jointly analyzes endpoint identifiability, shared server connectivity, symmetry in call duration and traffic volume, and service type indicators to derive correlation artifacts for matching. The framework is instantiated and evaluated on WhatsApp, Facebook Messenger, and Snapchat across diverse user behavior scenarios and commonly deployed network settings. Experimental results show that the method reliably links caller and callee flows, revealing edges in users’ social graphs without decrypting any packets. Under typical data retention regimes, these findings indicate that metadata-based correlation provides a practical basis for deanonymization and represents a persistent privacy risk for users of IM calling. Full article

Permissionless blockchains have evolved beyond cryptocurrency into foundations for Web3 applications, decentralized finance (DeFi), and digital asset ownership, yet this rapid expansion has intensified privacy vulnerabilities. This study provides a comprehensive review of recent trends, emerging privacy threats, and mitigation strategies in permissionless blockchain ecosystems. We examine six developments reshaping the landscape: meme coin proliferation on high-throughput networks, real-world asset tokenization linking on-chain activity to regulated identities, perpetual derivatives exposing trading strategies, institutional adoption concentrating holdings under regulatory oversight, prediction markets creating permanent records of beliefs, and blockchain–AI integration enabling both privacy-preserving analytics and advanced deanonymization. Through this work and forensic analysis of documented incidents, we analyze seven critical privacy threats grounded in verifiable 2024–2025 transaction data: dust attacks, private key management failures, transaction linking, remote procedure call exposure, maximal extractable value extraction, signature hijacking, and smart contract vulnerabilities. Blockchain exploits reached $2.36 billion in 2024 and $2.47 billion in the first half of 2025, with over 80% attributed to compromised private keys and signature vulnerabilities. We evaluate privacy-enhancing technologies, including zero-knowledge proofs, ring signatures, and stealth addresses, identifying the gap between academic proposals and production deployment. We further propose a Secure Development Lifecycle framework incorporating measurable security controls validated against incident data. This work bridges the disconnect between privacy research and industrial practice by synthesizing current trends, providing insights, documenting real-world threats with forensic evidence, and providing actionable insights for both researchers advancing privacy-preserving techniques and developers building secure blockchain applications. Full article

The image of the Virgin Mary appears with increasing frequency in written sources from the 12th and 13th centuries compared to earlier periods. Three major works produced by four eyewitness authors of the Albigensian Crusade (Historia Albigensis, Chronica, and Canso de la Crozada) reflect on and respond to this popular theme. These sources focus on the Albigensian Crusade against heretical groups, particularly the Cathars, and employ the Virgin Mary motif for various purposes. The Virgin Mary is presented as a Catholic model for women drawn to Catharism (a movement in which female spiritual leadership was also present) as a divine protector of the just side in war and as a means of legitimizing the authors’ claims. While Mary appears sporadically in Peter of Vaux-de-Cernay’s Historia Albigensis, she is extensively invoked in the Canso by both William and his anonymous successor. In contrast, the image of the Virgin Mary is scarcely mentioned in Chronica, likely due to the narrative’s intended audience and objectives. This article aims to provide a comparative analysis of how the image of the Virgin Mary is utilized in these primary sources from the Albigensian Crusade and to offer a new perspective on the relationship between historical events and authors’ intentions, laying the groundwork for further research. Full article

The secondary use of electronic health records is essential for developing artificial intelligence-based clinical decision support systems. However, even after direct identifiers are removed, de-identified electronic health records remain vulnerable to re-identification, membership inference attacks, and model extraction attacks. This study examined the balance between privacy protection and model utility by evaluating de-identification strategies and differentially private learning in large-scale electronic health records. De-identified records from a tertiary medical center were analyzed and compared with three strategies—baseline generalization, enhanced generalization, and enhanced generalization with suppression—together with differentially private stochastic gradient descent. Privacy risks were assessed through k-anonymity distributions, membership inference attacks, and model extraction attacks. Model performance was evaluated using standard predictive metrics, and privacy budgets were estimated for differentially private stochastic gradient descent. Enhanced generalization with suppression consistently improved k-anonymity distributions by reducing small, high-risk classes. Membership inference attacks remained at the chance level under all conditions, indicating that patient participation could not be inferred. Model extraction attacks closely replicated victim model outputs under baseline training but were substantially curtailed once differentially private stochastic gradient descent was applied. Notably, privacy-preserving learning maintained clinically relevant performance while mitigating privacy risks. Combining suppression with differentially private stochastic gradient descent reduced re-identification risk and markedly limited model extraction while sustaining predictive accuracy. These findings provide empirical evidence that a privacy–utility balance is achievable in clinical applications. Full article