arXiv subjects

Jiamin Zhang

Publications and source records attributed to Jiamin Zhang.

4 recordsLinked to original sources

arXiv2026

NTIRE 2026 Challenge on Robust AI-Generated Image Detection in the Wild

This paper presents an overview of the NTIRE 2026 Challenge on Robust AI-Generated Image Detection in the Wild, held in conjunction with the NTIRE workshop at CVPR 2026. The goal of this challenge was to develop detection models capable of distinguishing real images from generated ones in realistic scenarios: the images are often transformed (cropped, resized, compressed, blurred) for practical usage, and therefore, the detection models should be robust to such transformations. The challenge is based on a novel dataset consisting of 108,750 real and 185,750 AI-generated images from 42 generators comprising a large variety of open-source and closed-source models of various architectures, augmented with 36 image transformations. Methods were evaluated using ROC AUC on the full test set, including both transformed and untransformed images. A total of 511 participants registered, with 20 teams submitting valid final solutions. This report provides a comprehensive overview of the challenge, describes the proposed solutions, and can be used as a valuable reference for researchers and practitioners in increasing the robustness of the detection models to real-world transformations.

cs.CV↗

arXiv2026

FeatDistill: A Feature Distillation Enhanced Multi-Expert Ensemble Framework for Robust AI-generated Image Detection

The rapid iteration and widespread dissemination of deepfake technology have posed severe challenges to information security, making robust and generalizable detection of AI-generated forged images increasingly important. In this paper, we propose FeatDistill, an AI-generated image detection framework that integrates feature distillation with a multi-expert ensemble, developed for the NTIRE Challenge on Robust AI-Generated Image Detection in the Wild. The framework explicitly targets three practical bottlenecks in real-world forensics: degradation interference, insufficient feature representation, and limited generalization. Concretely, we build a four-backbone Vision Transformer (ViT) ensemble composed of CLIP and SigLIP variants to capture complementary forensic cues. To improve data coverage, we expand the training set and introduce comprehensive degradation modeling, which exposes the detector to diverse quality variations and synthesis artifacts commonly encountered in unconstrained scenarios. We further adopt a two-stage training paradigm: the model is first optimized with a standard binary classification objective, then refined by dense feature-level self-distillation for representation alignment. This design effectively mitigates overfitting and enhances semantic consistency of learned features. At inference time, the final prediction is obtained by averaging the probabilities from four independently trained experts, yielding stable and reliable decisions across unseen generators and complex degradations. Despite the ensemble design, the framework remains efficient, requiring only about 10 GB peak GPU memory. Extensive evaluations in the NTIRE challenge setting demonstrate that FeatDistill achieves strong robustness and generalization under diverse ``in-the-wild'' conditions, offering an effective and practical solution for real-world deepfake image detection.

cs.CV↗

arXiv2025

PoisonCatcher: Revealing and Identifying LDP Poisoning Attacks in IIoT

Local Differential Privacy (LDP), a robust privacy-protection model, is widely adopted in the Industrial Internet of Things (IIoT) due to its lightweight, decentralized, and scalable. However, its perturbation-based privacy-protection mechanism hinders distinguishing between any two data, thereby facilitating LDP poisoning attacks. The exposed physical-layer vulnerabilities and resource-constrained prevalent at the IIoT edge not only facilitate such attacks but also render existing LDP poisoning defenses, all of which are deployed at the edge and rely on ample resources, impractical. This work proposes a LDP poisoning defense for IIoT in the resource-rich aggregator. We first reveal key poisoning attack modes occurring within the LDP-utilized IIoT data-collection process, detailing how IIoT vulnerabilities enable attacks, and then formulate a general attack model and derive the poisoned data's indistinguishability. This work subsequently analyzes the poisoning impacts on aggregated data based on industrial process correlation, revealing the distortion of statistical query results' temporal similarity and the resulting disruption of inter-attribute correlation, and uncovering the intriguing paradox that adversaries' attempts to stabilize their poisoning actions for stealth are difficult to maintain. Given these findings, we propose PoisonCatcher, a solution for identifying poisoned data, which includes time-series detectors based on temporal similarity, attribute correlation, and pattern stability metrics to detect poisoned attributes, and a latent-bias feature miner for identifying poisons. Experiments on the real-world dataset indicate that PoisonCatcher successfully identifies poisoned data, demonstrating robust identification capabilities with F2 scores above 90.7\% under various attack settings.

cs.CR↗

arXiv2022

Preparation for future active learning

It is well documented that students sometimes resist active learning techniques. A recent study showed how students believed that they learned less in active learning classrooms than they learned in lectures, even though they learned more. In this article, we describe a method for introducing active learning methods to college students that is based on a preparation for future learning approach. The students who received this introduction to active learning appeared to be more receptive to group work in the classroom than students who started the course with an explanation of the reasons and values of active learning.

physics.ed-ph↗