StyleID: Identity Disentanglement for Anonymizing Faces

Le, Minh-Ha; Carlsson, Niklas

doi:10.56553/popets-2023-0016

Cited by 6 publications

(7 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We included as many works as we could find which appeared at USENIX Security, Privacy Enhancing Technologies Symposium (PETs), and Data and Applications Security and Privacy. As a recent work [38] of 2023 testifies to the persistence of the said methodological flaws to this day. An expanded survey of the current methodology can be found in the Appendix A.…”

Section: State-of-the-art Evaluation Methods For the Anonymization Of...mentioning

confidence: 98%

A False Sense of Privacy: Towards a Reliable Evaluation Methodology for the Anonymization of Biometric Data

Hanisch,

Todt,

Patino

et al. 2024

PoPETs

View full text Add to dashboard Cite

Biometric data contains distinctive human traits such as facial features or gait patterns. The use of biometric data permits an individuation so exact that the data is utilized effectively in identification and authentication systems. But for this same reason, privacy protections become indispensably necessary. Privacy protection is extensively afforded by the technique of anonymization. Anonymization techniques protect sensitive personal data from biometrics by obfuscating or removing information that allows linking records to the generating individuals, to achieve high levels of anonymity. However, our understanding and possibility to develop effective anonymization relies, in equal parts, on the effectiveness of the methods employed to evaluate anonymization performance. In this paper, we assess the state-of-the-art methods used to evaluate the performance of anonymization techniques for facial images and for gait patterns. We demonstrate that the state-of-the-art evaluation methods have serious and frequent shortcomings. In particular, we find that the underlying assumptions of the state-of-the-art are quite unwarranted. State-of-the-art methods generally assume a difficult recognition scenario and thus a weak adversary. However, that assumption causes state-of-the-art evaluations to grossly overestimate the performance of the anonymization. Therefore, we propose a strong adversary which is aware of the anonymization in place. This adversary model implements an appropriate measure of anonymization performance. We improve the selection process for the evaluation dataset, and we reduce the numbers of identities contained in the dataset while ensuring that these identities remain easily distinguishable from one another. Our novel evaluation methodology surpasses the state-of-the-art because we measure worst-case performance and so deliver a highly reliable evaluation of biometric anonymization techniques.

show abstract

Section: State-of-the-art Evaluation Methods For the Anonymization Of...mentioning

confidence: 98%

A False Sense of Privacy: Towards a Reliable Evaluation Methodology for the Anonymization of Biometric Data

Hanisch,

Todt,

Patino

et al. 2024

PoPETs

View full text Add to dashboard Cite

show abstract

“…The notable aspect of StyleGAN is its disentangled latent space which provides fine-grained control over image generation. This feature is explored and utilized by various researchers, such as Härkönen et al [81], Wu et al [210], Shen et al [176], and Le et al [114], to manipulate specific attributes within StyleGAN's latent space, and hence, modify corresponding features in generated images while maintaining realism and quality.…”

Section: Figure 22: Generative Adversarial Networkmentioning

confidence: 99%

“…In paper [114], we introduce a novel framework for the anonymization of facial image datasets, emphasizing the preservation of non-identity features within the anonymization process. Unlike traditional approaches that manipulate pixel space for de-identification, our methodology projects images into a GAN latent space.…”

Section: Ourmentioning

confidence: 99%

“…It is important to recognize that data poisoning represents just one facet of a multifaceted approach to protecting facial privacy. Our frameworks, including AnonFACES [115], StyleID [114], StyleAdv [113], and Diff-Private [111], alongside evasion techniques and early protection methods, provide a comprehensive toolbox against FRS. The following paragraphs outline several reasons why the battle for privacy, despite these challenges, is far from lost and why efforts to develop and refine a wide range of privacy protection strategies remain a critical component of the broader privacy protection toolkit.…”

Section: Diffprivate:contributionsmentioning

confidence: 99%

See 1 more Smart Citation

Beyond Recognition : Privacy Protections in a Surveilled World

Minh-Ha

2024

Linköping Studies in Science and Technology. Dissertations

View full text Add to dashboard Cite

Våra ansikten är unika och är en integrerad del av vår identitet. Med smartphones och sociala medier finns det miljarder ansiktsbilder online varje dag. Denna data driver ansiktsigenkänningsteknik, som används för allt från att låsa upp telefoner till omfattande övervakning. I takt med att ansiktsigenkänningstekniken utvecklas, ökar också dess förmåga att identifiera oss -ofta utan vårt samtycke. Denna verklighet leder oss till en avgörande fråga: Hur kan vi skydda vår integritet utan att offra fördelarna med den digitala åldern?Vår forskning antar denna utmaning genom att undersöka hur ansiktsigenkänningsteknik fungerar, dess fördelar för säkerhet och bekvämlighet, samt de integritetsrisker den medför. Vi har utvecklat innovativa strategier för att skydda vår ansiktsdata från icke-samtyckt övervakning samtidigt som vi behåller nyttan med systemen inom legitima områden som behöver anonymiserade data för samhällets bästa, såsom sjukvård, självkörande fordon och maskininlärning. För att navigera dessa utmaningar har vi introducerat banbrytande verktyg och metoder för anonymisering, såsom AnonFACES, Sty-leID och utforskningen av sårbarheter i FRS genom IdDecoder.Föreställ dig en värld där du kan dela bilder online utan rädsla för att de utnyttjas av övervakningssystem. För att ta itu med dessa problem har vi skapat nya verktyg och tekniker, såsom DiffPrivate och StyleAdv. Dessa innovationer gör ansiktsbilder "osynliga" för övervakningssystem utan att ändra hur de ser ut för mänskliga ögon. Dessa verktyg är konstruerade för att vara offentligt tillgängliga samtidigt som de ger oss möjligheten att effektivt skydda vår integritet.Vårt arbete banar väg för en framtid där tekniken ger oss möjlighet att njuta av de stora möjligheterna med den digitala världen samtidigt som vi upprätthåller vår grundläggande rätt till integritet. Det handlar om att upprätthålla en harmonisk balans mellan innovation och integritet, och säkerställa att våra unika identiteter, som representeras av våra ansikten, förblir uteslutande våra i en värld som ständigt söker märka och kategorisera oss.

show abstract

“…Other researchers have also proposed approaches to alter the facial attributes in images using generative models [34], [35], [36], [37], [38]. Researchers have also found ways to disentangle the identity of a person from other facial attributes of the image [39], [40].…”

Section: Synthetic Data Generationmentioning

confidence: 99%

Zero-shot demographically unbiased image generation from an existing biased StyleGAN

Jain,

Dholakia,

Memon

et al. 2023

Preprint

View full text Add to dashboard Cite

<p>Face recognition systems have made significant strides thanks to data-heavy deep learning models, but these models rely on large privacy-sensitive datasets. Recent work in facial analysis and recognition have thus started making use of synthetic datasets generated from GANs and diffusion based generative models. These models, however, lack fairness in terms of demographic representation and can introduce the same biases in the trained downstream tasks. This can have serious societal and security implications. To address this issue, we propose a methodology that generates unbiased data from a biased generative model using an evolutionary algorithm. We show results for StyleGAN2 model trained on the Flicker Faces High Quality dataset to generate data for singular and combinations of demographic attributes such as \textit{Black and Woman}. We generate a large racially balanced dataset of 13.5 million images, and show that it boosts the performance of facial recognition and analysis systems whilst reducing their biases. </p>

show abstract

StyleID: Identity Disentanglement for Anonymizing Faces

Cited by 6 publications

References 30 publications

A False Sense of Privacy: Towards a Reliable Evaluation Methodology for the Anonymization of Biometric Data

A False Sense of Privacy: Towards a Reliable Evaluation Methodology for the Anonymization of Biometric Data

Beyond Recognition : Privacy Protections in a Surveilled World

Zero-shot demographically unbiased image generation from an existing biased StyleGAN

Contact Info

Product

Resources

About