DeepFakes: a New Threat to Face Recognition? Assessment and Detection

Sébastien Marcel
Idiap, 2018
Rapport ou document de travail

Résumé

It is becoming increasingly easy to automatically replace a face of one person in a video with the face of another person by using a pre-trained generative adversarial network (GAN). Recent public scandals, e.g., the faces of celebrities being swapped onto pornographic videos, call for automated ways to detect these Deepfake videos. To help developing such methods, in this paper, we present the first publicly available set of Deepfake videos generated from videos of VidTIMIT database. We used open source software based on GANs to create the Deepfakes, and we emphasize that training and blending parameters can significantly impact the quality of the resulted videos. To demonstrate this impact, we generated videos with low and high visual quality (320 videos each) using differently tuned parameter sets. We showed that the state of the art face recognition systems based on VGG and Facenet neural networks are vulnerable to Deepfake videos, with 85.62% and 95.00% false acceptance rates respectively, which means methods for detecting Deepfake videos are necessary. By considering several baseline approaches, we found that audio-visual approach based on lip-sync inconsistency detection was not able to distinguish Deepfake videos. The best performing method, which is based on visual quality metrics and is often used in presentation attack detection domain, resulted in 8.97% equal error rate on high quality Deepfakes. Our experiments demonstrate that GAN-generated Deepfake videos are challenging for both face recognition systems and existing detection methods, and the further development of face swapping technology will make it even more so.

Official source

https://infoscience.epfl.ch/record/263134?ln=fr

À propos de ce résultat

Cette page est générée automatiquement et peut contenir des informations qui ne sont pas correctes, complètes, à jour ou pertinentes par rapport à votre recherche. Il en va de même pour toutes les autres pages de ce site. Veillez à vérifier les informations auprès des sources officielles de l'EPFL.

Concepts associés

Chargement

Publications associées

Chargement

Sébastien Marcel
Idiap, 2018
Rapport ou document de travail

Résumé

Official source

https://infoscience.epfl.ch/record/263134?ln=fr

À propos de ce résultat

Concepts associés (12)

Concepts associés

Chargement

Publications associées (19)

Publications associées

Chargement

Makeup is a simple and easy instrument that can alter the appearance of a person’s face, and hence, create a presentation attack on face recognition (FR) systems. These attacks, especially the ones mimicking ageing, are difficult to detect due to their close resemblance with genuine (non-makeup) appearances. Makeups can also degrade the performance of recognition systems and of various algorithms that use human face as an input. The detection of facial makeups is an effective prohibitory measure to minimize these problems. This work proposes a deep learning-based presentation attack detection (PAD) method to identify facial makeups. We propose the use of a convolutional neural network (CNN) to extract features that can distinguish between presentations with age-induced facial makeups (attacks), and those without makeup (bona-fide). These feature descriptors, based on shape and texture cues, are constructed from multiple intermediate layers of a CNN. We introduce a new dataset AIM (Age Induced Makeups) consisting of 200+ video presentations of old-age makeups and bona-fide, each. Our experiments indicate makeups in AIM result in 14% decrease in the median matching scores of a recent CNN-based FR system. We demonstrate accuracy of the proposed PAD method where 93% presentations in the AIM dataset are correctly classified. In additional testing, it also outperforms existing methods of detection of generic makeups. A simple score-level fusion, performed on the classification scores of shape- and texture-based features, can further improve the accuracy of the proposed makeup detector.

2019

Chargement

On Performance / Robustness / Complexity Trade-Offs in Face Verification

Samy Bengio

In much of the literature devoted to face recognition, experiments are performed with controlled images (e.g. manual face localization, controlled lighting, background and pose); however, a practical recognition system has to be robust to more challenging conditions. In this paper we first evaluate, on the relatively difficult BANCA database, the performance, robustness and complexity of Gaussian Mixture Model (GMM), 1D- and pseudo-2D Hidden Markov Model (HMM) based systems, using both manual and automatic face localization. We also propose to extend the GMM approach through the use of local features with embedded positional information, increasing performance without sacrificing its low complexity. Experiments show that good performance on manually located faces is not necessarily indicative of good performance on automatically located faces (which are imperfectly located). The deciding factor is shown to be the degree of constraints placed on spatial relations between face parts. Methods which utilize rigid constraints have poor robustness compared to methods which have relaxed constraints. Furthermore, we show that while the pseudo-2D HMM approach has the best overall performance, classification time on current hardware makes it impractical. The best trade-off in terms of complexity, robustness and discrimination performance is achieved by the extended GMM approach.

IDIAP2004

Chargement

User Authentication via Adapted Statistical Models of Face Images

Samy Bengio

It has been previously demonstrated that systems based on local features and relatively complex statistical models, namely 1D Hidden Markov Models (HMMs) and pseudo-2D HMMs, are suitable for face recognition. Recently, a simpler statistical model, namely the Gaussian Mixture Model (GMM), was also shown to perform well. In much of the literature devoted to these models, the experiments were performed with controlled images (manual face localization, controlled lighting, background, pose, etc.); however, a practical recognition system has to be robust to more challenging conditions. In this article we evaluate, on the relatively difficult BANCA database, the performance, robustness and complexity of GMM and HMM based approaches, using both manual and automatic face localization. We extend the GMM approach through the use of local features with embedded positional information, increasing performance without sacrificing its low complexity. Furthermore, we show that the traditionally used Maximum Likelihood (ML) training approach has problems estimating robust model parameters when there is only a few training images available; considerably more precise models can be obtained through the use of Maximum a Posteriori (MAP) training. We also show that face recognition techniques which obtain good performance on manually located faces do not necessarily obtain good performance on automatically located faces, indicating that recognition techniques must be designed from the ground up to handle imperfect localization. Finally, we show that while the pseudo 2D HMM approach has the best overall performance, authentication time on current hardware makes it impractical; the best trade-off in terms of authentication time, robustness and discrimination performance is achieved by the extended GMM approach.

IDIAP2004

Chargement