Detecting deepfakes involving face-swaps presents a significant challenge, particularly in real-world scenarios where anyone can perform face-swapping with freely available tools and apps without any technical knowledge. Existing deepfake detection methods rely on facial landmarks or inconsistencies in pixel-level features and often struggle with face-swap deepfakes, where the source face is seamlessly blended into the target image or video. The prevalence of face-swap is evident in everyday life, where it is used to spread false information, damage reputations, manipulate political opinions, create non-consensual intimate deepfakes (NCID), and exploit children by enabling the creation of child sexual abuse material (CSAM). Even prominent public figures are not immune to its impact, with numerous deepfakes of them circulating widely across social media platforms. Another challenge faced by deepfake detection methods is the creation of datasets that encompass a wide range of variations, as training models require substantial amounts of data. This raises privacy concerns, particularly regarding the processing and storage of personal facial data, which could lead to unauthorized access or misuse. Our key idea is to identify these style discrepancies to detect face-swapped images effectively without accessing the real facial image. We perform comprehensive evaluations using multiple datasets and face-swapping methods, which showcases the effectiveness of SafeVision in detecting face-swap deepfakes across diverse scenarios. SafeVision offers a reliable and scalable solution for detecting face-swaps in a privacy preserving manner, making it particularly effective in challenging real-world applications. To the best of our knowledge, SafeVision is the first deepfake detection using style features while providing inherent privacy protection.

Language models trained with reinforcement learning (RL) can engage in reward hacking--the exploitation of unintended strategies for high reward--without revealing this behavior in their chain-of-thought reasoning. This makes the detection of reward hacking difficult, posing risks for high-stakes applications. We propose verbalization fine-tuning (VFT), a pre-RL fine-tuning intervention that trains models to explicitly acknowledge when they are influenced by prompt cues--hints which point to incorrect answers (e.g., "a Stanford professor thinks the answer is A"). To evaluate VFT, we subsequently train models with RL on environments where held-out prompt cues signal which incorrect answers will receive high reward, incentivizing models to exploit these cues instead of reasoning correctly. We measure how often models exploit these cues without verbalizing it. After RL, only 6% of the VFT-trained model's responses consist of undetected reward hacks. In comparison, when we perform RL without VFT, the rate of undetected reward hacks goes up to 88%; with a debiasing baseline intervention, this increases further to 99%. VFT achieves this by substantially increasing how often models verbalize the influence of cues, from 8% to 43% after VFT, and up to 94% after RL. Baselines remain low even after RL (11% and 1%). Our results show that teaching models to explicitly verbalize reward hacking behavior before RL significantly improves their detection, offering a practical path toward more transparent and safe AI systems.

A key challenge in security analysis is the manual evaluation of potential security weaknesses generated by static application security testing (SAST) tools. Numerous false positives (FPs) in these reports reduce the effectiveness of security analysis. We propose using Large Language Models (LLMs) to improve the assessment of SAST findings. We investigate the ability of LLMs to reduce FPs while trying to maintain a perfect true positive rate, using datasets extracted from the OWASP Benchmark (v1.2) and a real-world software project. Our results indicate that advanced prompting techniques, such as Chain-of-Thought and Self-Consistency, substantially improve FP detection. Notably, some LLMs identified approximately 62.5% of FPs in the OWASP Benchmark dataset without missing genuine weaknesses. Combining detections from different LLMs would increase this FP detection to approximately 78.9%. Additionally, we demonstrate our approach's generalizability using a real-world dataset covering five SAST tools, three programming languages, and infrastructure files. The best LLM detected 33.85% of all FPs without missing genuine weaknesses, while combining detections from different LLMs would increase this detection to 38.46%. Our findings highlight the potential of LLMs to complement traditional SAST tools, enhancing automation and reducing resources spent addressing false alarms.

The design of science-based policies to improve the sustainability of smallholder agriculture is challenged by a limited understanding of fundamental system properties, such as the spatial distribution of active cropland and field size. We integrate very high spatial resolution (1.5 m) Earth observation data and deep transfer learning to derive crop field delineations in complex agricultural systems at the national scale, while maintaining minimum reference data requirements and enhancing transferability. We provide the first national-level dataset of 21 million individual fields for Mozambique (covering ~800,000 km2) for 2023. Our maps separate active cropland from non-agricultural land use with an overall accuracy of 93% and balanced omission and commission errors. Field-level spatial agreement reached median intersection over union (IoU) scores of 0.81, advancing the state-of-the-art in large-area field delineation in complex smallholder systems. The active cropland maps capture fragmented rural regions with low cropland shares not yet identified in global land cover or cropland maps. These regions are mostly located in agricultural frontier regions which host 7-9% of the Mozambican population. Field size in Mozambique is very low overall, with half of the fields being smaller than 0.16 ha, and 83% smaller than 0.5 ha. Mean field size at aggregate spatial resolution (0.05°) is 0.32 ha, but it varies strongly across gradients of accessibility, population density, and net forest cover change. This variation reflects a diverse set of actors, ranging from semi-subsistence smallholder farms to medium-scale commercial farming, and large-scale farming operations. Our results highlight that field size is a key indicator relating to socio-economic and environmental outcomes of agriculture (e.g., food production, livelihoods, deforestation, biodiversity), as well as their trade-offs.

--Detecting DNS tunneling is a significant challenge in cybersecurity due to its capacity to hide harmful actions within DNS traffic that appears to be normal and legitimate. Traditional detection methods based on rule-based approaches or signature matching are often insufficient to accurately identify such covert communication channels. This paper addresses the necessity of machine learning methods for effective DNS tunneling detection. We propose a novel approach to detect DNS tunneling. Through the combination of advanced machine learning algorithms and the analysis of various features extracted from DNS traffic, our aim is to provide an accurate DNS tunneling detection model. A. About the Subject The Domain Name System (DNS) is a hierarchical and decentralized naming system crucial for internet functionality [1]. As a core component of internet infrastructure, DNS is used in nearly every online transaction, making it a prime target for a variety of cyber threats. Due to its foundational role and widespread trust, DNS is vulnerable to several types of attacks, threat landscape can be seen in [2], such as cache poisoning, amplification and DoS attacks, and phishing attacks. These vulnerabilities offer attackers multiple possibilities to disrupt or manipulate internet traffic.

In the dynamic realm of online recruitment, accurate job classification is paramount for optimizing job recommendation systems, search rankings, and labor market analyses. As job markets evolve, the increasing complexity of job titles and descriptions necessitates sophisticated models that can effectively leverage intricate relationships within job data. Traditional text classification methods often fall short, particularly due to their inability to fully utilize the hierarchical nature of industry categories. To address these limitations, we propose a novel representation learning and classification model that embeds jobs and hierarchical industry categories into a latent embedding space. Our model integrates the Standard Occupational Classification (SOC) system and an in-house hierarchical taxonomy, Carotene, to capture both graph and hierarchical relationships, thereby improving classification accuracy. By embedding hierarchical industry categories into a shared latent space, we tackle cold start issues and enhance the dynamic matching of candidates to job opportunities. Extensive experimentation on a large-scale dataset of job postings demonstrates the model's superior ability to leverage hierarchical structures and rich semantic features, significantly outperforming existing methods. This research provides a robust framework for improving job classification accuracy, supporting more informed decision-making in the recruitment industry.

This paper aims to propose a novel machine learning (ML) approach incorporating Homomorphic Encryption (HE) to address privacy limitations in Unmanned Aerial Vehicles (UAV)-based face detection. Due to challenges related to distance, altitude, and face orientation, high-resolution imagery and sophisticated neural networks enable accurate face recognition in dynamic environments. However, privacy concerns arise from the extensive surveillance capabilities of UAVs. To resolve this issue, we propose a novel framework that integrates HE with advanced neural networks to secure facial data throughout the inference phase. This method ensures that facial data remains secure with minimal impact on detection accuracy. Specifically, the proposed system leverages the Cheon-Kim-Kim-Song (CKKS) scheme to perform computations directly on encrypted data, optimizing computational efficiency and security. Furthermore, we develop an effective data encoding method specifically designed to preprocess the raw facial data into CKKS form in a Single-Instruction-Multiple-Data (SIMD) manner. Building on this, we design a secure inference algorithm to compute on ciphertext without needing decryption. This approach not only protects data privacy during the processing of facial data but also enhances the efficiency of UAV-based face detection systems. Experimental results demonstrate that our method effectively balances privacy protection and detection performance, making it a viable solution for UAV-based secure face detection. Significantly, our approach (while maintaining data confidentially with HE encryption) can still achieve an accuracy of less than 1% compared to the benchmark without using encryption.

Diabetic retinopathy (DR) is a leading cause of vision loss in working-age adults. While screening reduces the risk of blindness, traditional imaging is often costly and inaccessible. Artificial intelligence (AI) algorithms present a scalable diagnostic solution, but concerns regarding fairness and generalization persist. This work evaluates the fairness and performance of image-trained models in DR prediction, as well as the impact of disentanglement as a bias mitigation technique, using the diverse mBRSET fundus dataset. Three models, ConvNeXt V2, DINOv2, and Swin V2, were trained on macula images to predict DR and sensitive attributes (SAs) (e.g., age and gender/sex). Fairness was assessed between subgroups of SAs, and disentanglement was applied to reduce bias. All models achieved high DR prediction performance in diagnosing (up to 94% AUROC) and could reasonably predict age and gender/sex (91% and 77% AUROC, respectively). Fairness assessment suggests disparities, such as a 10% AUROC gap between age groups in DINOv2. Disentangling SAs from DR prediction had varying results, depending on the model selected. Disentanglement improved DINOv2 performance (2% AUROC gain), but led to performance drops in ConvNeXt V2 and Swin V2 (7% and 3%, respectively). These findings highlight the complexity of disentangling fine-grained features in fundus imaging and emphasize the importance of fairness in medical imaging AI to ensure equitable and reliable healthcare solutions.

Driving trajectory data remains vulnerable to privacy breaches despite existing mitigation measures. Traditional methods for detecting driving trajectories typically rely on map-matching the path using Global Positioning System (GPS) data, which is susceptible to GPS data outage. This paper introduces CAN-Trace, a novel privacy attack mechanism that leverages Controller Area Network (CAN) messages to uncover driving trajectories, posing a significant risk to drivers' long-term privacy. A new trajectory reconstruction algorithm is proposed to transform the CAN messages, specifically vehicle speed and accelerator pedal position, into weighted graphs accommodating various driving statuses. CAN-Trace identifies driving trajectories using graph-matching algorithms applied to the created graphs in comparison to road networks. We also design a new metric to evaluate matched candidates, which allows for potential data gaps and matching inaccuracies. Empirical validation under various real-world conditions, encompassing different vehicles and driving regions, demonstrates the efficacy of CAN-Trace: it achieves an attack success rate of up to 90.59% in the urban region, and 99.41% in the suburban region.

This study explores the optimization of the DRAGON Longformer base model for clinical text classification, specifically targeting the binary classification of medical case descriptions. A dataset of 500 clinical cases containing structured medical observations was used, with 400 cases for training and 100 for validation. Enhancements to the pre - trained joeranbosma/dragon - longformer - base - mixed - domain model included hyperparameter tuning, domain - specific preprocessing, and architectural adjustments. Key modifications involved increasing sequence length from 512 to 1024 tokens, adjusting learning rates from 1e - 05 to 5e - 06, extending training epochs from 5 to 8, and incorporating specialized medical terminology. The optimized model achieved notable performance gains: accuracy improved from 72.0% to 85.2%, precision from 68.0% to 84.1%, recall from 75.0% to 86.3%, and F1 - score from 71.0% to 85.2%. Statistical analysis confirmed the significance of these improvements (p < .001). The model demonstrated enhanced capability in interpreting medical terminology, anatomical measurements, and clinical observations. These findings contribute to domain - specific language model research and offer practical implications for clinical natural language processing applications. The optimized model ' s strong performance across diverse medical conditions underscores its potential for broad use in healthcare settings. Enhancing Clinical Text Classification via Fine - Tuned DRAGON Longformer Models Introduction Natural language processing (NLP) in healthcare has continued to advance rapidly, revolutionizing the ability to analyze clinical texts and automate the extraction of valuable insights from massive amounts of medical documentation (Khurana, Koli, Khatter, & Singh, 2023). Over the past few years, large language models (LLMs) have emerged as powerful tools for gaining insight from and processing clinical narratives, creating capabilities that have never been seen before in medical text classification, entity recognition, and clinical decision support (Wang et al., 2018). The DRAGON (Deep Representation Analysis for General - domain Ontology Networks) framework was a specialized version of medical text processing out of all these models (Bosma et al., 2025). Beltagy, Peters, and Cohan (2020) state that the DRAGON longformer model, built on top of the Longformer architecture, addresses the quadratic computational complexity issue of traditional transformer models by processing long sequences.