11-19 更新 实时
首页/ 算力/ 正文

物理启发AI通过分析颜色与亮度模式识别伪造人脸

算力 来源:TechXplore 2026-11-14 00:00 0 阅读 机器翻译
物理启发AI通过分析颜色与亮度模式识别伪造人脸

由 Gaby Clark 编辑,Robert Egan 审阅

本文已按照 Science X 的编辑流程与政策进行审阅。编辑在确保内容可信度的同时,重点标注了以下属性:

添加到首选来源

RGB 图像及对应的 FDIP 图(一种类似散度的响应、将梯度幅值解释为类似速度的度量,以及一种类似动能的响应),展示了如何揭示超出 RGB 外观的结构信息(原始输入图像经 Deepfake and Real Images (DFRI) 数据集许可改编,Manjil Karki (2020),https://www.kaggle.com/datasets/manjilkarki/deepfake-and-real-images)。图片来源:Frontiers in Imaging (2026)。DOI: 10.3389/fimag.2026.1935077

基尔大学(Keele University)的研究人员开发出一种受物理学启发的 AI 模型,在区分真实人脸图像与 AI 生成或篡改图像的测试中取得了高准确率。

该研究团队由基尔大学计算机科学与数学学院的博士生 Saman Maghsoudi 领导,他们开发出一种工具,通过分析图像中的颜色信息和亮度模式,帮助区分真实人脸图像与 AI 生成或篡改的图像。

在生成式 AI 时代,制作人脸的合成照片或视频(即所谓“深度伪造”)变得愈发容易,这意味着可靠识别真实或伪造图像的能力,对于数字媒体等领域的信任、安全和真实性以及打击身份盗用正变得越来越重要。

亮度模式提供了额外线索

由研究团队(包括 Dr. Sangeeta Sangeeta 和 Dr. Nadia Kanwal)提出的模型名为 FDIP-NET。它将传统的彩色图像信息与三项额外测量值相结合,这些测量值源自图像中亮度变化的方式。这些测量受到流体力学数学概念的启发,用于描述图像的局部结构,而并不对实际的流体运动进行建模。

该模型分析图像亮度的颜色变化与空间变化,而非追踪人的动作或随时间发生的变化。该框架还会校准预测置信度,并考察检测误差与模糊、亮度和对比度等图像属性之间的关联。研究人员使用公开可获取的人脸图像数据集 DFRI、DF-wild 和 RFF-140K 对其进行了评估。

在单个数据集内保持高准确率

在实验中,FDIP-NET 在 DFRI、DF-wild 和 RFF-140K 上的平均测试准确率分别为 98.654%、98.952% 和 99.669%。这些结果是对三次训练运行取平均得出的,每个数据集内的模型均单独进行训练和测试。在使用官方 FaceForensics++ c23 的训练、验证和测试划分进行的额外评估中,该模型在单个视频帧上取得了 91% 的平均准确率。

该研究成果发表在 Frontiers in Imaging 上,凸显了将传统彩色图像特征与源自图像亮度的结构信息相结合以支持深度伪造检测的潜力。其收益因数据集和评估指标而异,而单独的测试表明,在陌生数据集上实现可靠检测仍是未来研究面临的一项挑战。

第一作者 Maghsoudi 表示:“随着 AI 生成的人脸越来越逼真,评估图像真实性对于数字媒体的可信度、以及应对身份冒用和虚假信息的工作都愈发重要。我们的研究结果表明,图像中的亮度模式可以补充传统的 AI 特征,同时也凸显了为何预测置信度和检测错误应当与准确率一样受到关注。这为进一步研究奠定了基础,以开发在陌生图像和真实世界条件下更可靠运行的检测器。”

Saman Maghsoudi 等,《FDIP-NET:一种受物理学启发的深度伪造图像检测器,利用 FDIP 亮度动力学与校准 CNN》,Frontiers in Imaging(2026)。DOI: 10.3389/fimag.2026.1935077

由基尔大学提供

这篇报道背后是谁?

英语硕士,自 2021 年起担任文字编辑,拥有高等教育和健康领域内容经验。致力于可信赖的科学新闻。完整简介 →

数学生物学学士,创意写作硕士。游历广泛,对科学与语言有独到见解。完整简介 →

引用:受物理学启发的 AI 通过分析颜色和亮度模式检测假脸(2026 年 11 月 13 日),2026 年 11 月 13 日检索自 https://techxplore.com/news/2026-11-physics-ai-fake-brightness-patterns.html

本文件受版权保护。除出于私人学习或研究目的的合理使用外,未经书面许可,不得复制任何部分。内容仅供参考。

新软件将帮助应对深度伪造图像对个人安全的威胁

面部运动分析以超过 95% 的准确率检测深度伪造视频

新工具利用 AI“指纹”检测被篡改的照片和视频

研究提出检测深度伪造的新方法

简短培训帮助人们在打击深度伪造欺诈中识别 AI 生成的人脸

高分辨率仿生眼将彩色成像与更高效的运动检测相结合

隐藏的守门人:微小岩石孔隙中的水如何塑造地下储能

可变形的生物纱线为可穿戴技术编织新可能

可重构 MEMS 开关或许能使芯片封装布线避开 X 射线逆向工程

研究人员破解了一个关于辐射的物理学难题,这有望催生更坚固耐用的电子器件

早期测试中,手势和振动腰带使无人机无需视觉反馈即可受控

获取即时摘要文本(Gist)

FDIP-NET 将颜色特征与受流体动力学启发的空间亮度变化度量相结合,以检测伪造人脸图像。在数据集内测试中,平均准确率达到 98.65–99.67%,在 FaceForensics++ 视频帧上达到 91%。在陌生数据集上实现可靠检测仍具挑战性。

本摘要使用 LLM 自动生成。完整免责声明

译文由英文原文机器翻译,术语按行业惯例统一,可在文末展开原文核对。

查看英文原文(Physics-inspired AI detects fake faces by analyzing color and brightness patterns)

edited by Gaby Clark, reviewed by Robert Egan

This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility:

Add to Preferred Sources

Keele researchers develop physics-inspired deepfake detection tool RGB image and corresponding FDIP maps (a divergence-like response, gradient magnitude interpreted as a speed-like measure, and a kinetic-energy-like response) illustrating how structural information is revealed beyond RGB appearance (original input image adapted with permission from the Deepfake and Real Images (DFRI) dataset, Manjil Karki (2020), https://www.kaggle.com/datasets/manjilkarki/deepfake-and-real-images). Credit: Frontiers in Imaging (2026). DOI: 10.3389/fimag.2026.1935077

Researchers at Keele University have developed a physics-inspired AI model that achieved high accuracy in tests distinguishing genuine facial images from AI-generated or manipulated images.

The research team, led by Ph.D. candidate Saman Maghsoudi from Keele's School of Computer Science and Mathematics, has created a tool that analyzes color information and patterns of brightness within an image to help distinguish genuine facial images from AI-generated or manipulated ones.

Creating synthetic photographs or videos of faces, known as "deepfakes," has become even easier in the age of generative AI, meaning the ability to reliably spot genuine or fake images is becoming increasingly important for trust, security and authenticity in things like digital media and combating identity theft.

Brightness patterns offer additional clues

The model proposed by the research team, including Dr. Sangeeta Sangeeta and Dr. Nadia Kanwal, is known as FDIP-NET. It combines conventional color-image information with three additional measurements derived from how brightness varies across an image. These measurements are inspired by mathematical concepts from fluid dynamics and describe local image structure without modeling physical fluid motion.

The model analyzes color and spatial variations in image brightness, rather than tracking a person's movement or changes over time. The framework also calibrates prediction confidence and examines how detection errors are associated with image properties such as blur, brightness and contrast. The researchers evaluated it using publicly available facial-image datasets known as DFRI, DF-wild and RFF-140K.

High accuracy within individual datasets

In their experiments, FDIP-NET achieved mean test accuracies of 98.654%, 98.952% and 99.669% on DFRI, DF-wild and RFF-140K, respectively. These results were averaged across three training runs, with models trained and tested separately within each dataset. In an additional evaluation using the official FaceForensics++ c23 training, validation and test splits, the model achieved 91% mean accuracy on individual video frames.

Published in Frontiers in Imaging, the findings highlight the potential of combining conventional color-image features with structural information derived from image brightness to support deepfake detection. The benefits varied across datasets and evaluation measures, and separate tests showed that reliable detection on unfamiliar datasets remains a challenge for future research.

Lead author Maghsoudi said, "As AI-generated faces become more convincing, assessing image authenticity is increasingly important for trust in digital media and efforts to tackle impersonation and misinformation. Our findings show how patterns of brightness within images can complement conventional AI features, while highlighting why prediction confidence and detection errors deserve attention alongside accuracy. This provides a foundation for further research toward detectors that work more reliably on unfamiliar images and under real-world conditions."

Saman Maghsoudi et al., FDIP-NET: a physics-inspired deepfake image detector leveraging FDIP luminance dynamics and calibrated CNNs, Frontiers in Imaging (2026). DOI: 10.3389/fimag.2026.1935077

Provided by Keele University

Who's behind this story?

Gaby Clark

MA in English, copy editor since 2021 with experience in higher education and health content. Dedicated to trustworthy science news. Full profile →

Bachelor's in mathematical biology, Master's in creative writing. Well-traveled with unique perspectives on science and language. Full profile →

Citation: Physics-inspired AI detects fake faces by analyzing color and brightness patterns (2026, November 13) retrieved 13 November 2026 from https://techxplore.com/news/2026-11-physics-ai-fake-brightness-patterns.html

This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.

New software will help combat deep fake image threats to personal security

Facial movement analysis detects deepfake videos with more than 95% accuracy

New tools use AI 'fingerprints' to detect altered photos, videos

Research introduces new approach for detecting deepfakes

Short training helps people spot AI faces in the battle against deepfake fraud

High-resolution bionic eye combines color imaging with more efficient motion detection

The hidden gatekeeper: How water in tiny rock pores shapes underground energy storage

Shape-shifting bioyarns weave new possibilities for wearable tech

Reconfigurable MEMS switches could hide chip package wiring from X-ray reverse engineering

Researchers solve a physics puzzle about radiation that could lead to tougher electronics

Hand gestures and vibrating belt enable drone control without visual feedback in early tests

Get Instant Summarized Text (Gist)

FDIP-NET combines color features with fluid-dynamics-inspired measures of spatial brightness variation to detect fake facial images. Mean accuracy reached 98.65–99.67% in within-dataset tests and 91% on FaceForensics++ video frames. Reliable detection on unfamiliar datasets remains challenging.

This summary was automatically generated using LLM. Full disclaimer

This site uses cookies to assist with navigation, analyse your use of our services, collect data for ads personalisation and provide content from third parties. By using our site, you acknowledge that you have read and understand our Privacy Policy and Terms of Use.

原文为英文,正文由自动抓取并保留原文。

本文由采集器自动同步

原文发布于 TechXplore(Physics-inspired AI detects fake faces by analyzing color and brightness patterns),标题、摘要与正文由机器翻译,术语按行业惯例统一,未经人工逐句校订。

阅读原文 ↗