Home Community The AI-Makeup Artist that Covers Your Identity: CLIP2Protect is an AI Model That Uses Text-Guided Makeup to Protect Facial Privacy

The AI-Makeup Artist that Covers Your Identity: CLIP2Protect is an AI Model That Uses Text-Guided Makeup to Protect Facial Privacy

0
The AI-Makeup Artist that Covers Your Identity: CLIP2Protect is an AI Model That Uses Text-Guided Makeup to Protect Facial Privacy

The 90s Sci-fi movies are stuffed with computers that show this rotating profile of an individual and display all sorts of data concerning the person. This face-recognition technology is predicted to be so advanced that no data about you’ll be able to stay hidden from the big-brother.

We cannot claim they were improper, unfortunately. Face recognition technology has witnessed significant advancements with the appearance of deep learning-based systems, revolutionizing various applications and industries. Whether this revolution was something good or bad is a subject for an additional post, but the fact is that our faces may be linked to a lot data about us in our world. On this case, privacy plays a vital role.

In response to those concerns, the research community has been actively exploring methods and techniques to develop facial privacy protection algorithms that may safeguard individuals against the potential risks related to face recognition systems.

[Sponsored] 🔥 Construct your personal brand with Taplio  🚀 The first all-in-one AI-powered tool to grow on LinkedIn. Create higher LinkedIn content 10x faster, schedule, analyze your stats & engage. Try it without cost!

The goal of facial privacy protection algorithms is to search out a balance between preserving a person’s privacy and maintaining the usability of their facial images. While the first objective is to guard individuals from unauthorized identification or tracking, it’s equally necessary to be certain that the protected images retain visual fidelity and resemblance to the unique faces in order that the system can’t be tricked with a fake face. 

Achieving this balance is difficult, particularly when using noise-based methods that overlay adversarial artifacts on the unique face image. Several approaches have been proposed to generate unrestricted adversarial examples, with adversarial makeup-based methods being the preferred ones for his or her ability to embed adversarial modifications in a more natural manner. Nevertheless, existing techniques suffer from limitations akin to makeup artifacts, dependence on reference images, the necessity for retraining for every goal identity, and a concentrate on impersonation reasonably than privacy preservation.

So, there may be a necessity for a reliable method to guard facial privacy, but existing ones suffer from obvious shortcomings. How can we solve this? Time to satisfy CLIP2Protect.

CLIP2Protect is a novel approach for safeguarding user facial privacy on online platforms. It involves trying to find adversarial latent codes in a low-dimensional manifold learned by a generative model. These latent codes may be used to generate high-quality face images that maintain a sensible face identity while deceiving black-box FR systems. 

A key component of CLIP2Protect is using textual prompts to facilitate adversarial makeup transfer, allowing the traversal of the generative model’s latent manifold to search out transferable adversarial latent codes. This system effectively hides attack information throughout the desired makeup style without requiring large makeup datasets or retraining for various goal identities. CLIP2Protect  also introduces an identity-preserving regularization technique to make sure the protected face images visually resemble the unique faces.

To make sure the naturalness and fidelity of the protected images, the seek for adversarial faces is constrained to remain near the clean image manifold learned by the generative model. This restriction helps mitigate the generation of artifacts or unrealistic features that might be easily detected by human observers or automated systems. Moreover, CLIP2Protect  focuses on optimizing only the identity-preserving latent codes within the latent space, ensuring that the protected faces retain the human-perceived identity of the person.

To introduce privacy-enhancing perturbations, CLIP2Protect  utilizes text prompts as guidance for generating makeup-like transformations. This approach offers greater flexibility to the user than reference image-based methods, because it allows for the specification of desired makeup styles and attributes through textual descriptions. By leveraging these textual prompts, the strategy can effectively embed privacy protection information within the makeup style without having a big makeup dataset or retraining for various goal identities.

Extensive experiments are conducted to judge the effectiveness of the CLIP2Protect  in face verification and identification scenarios. The outcomes show its efficacy against black-box FR models and online industrial facial recognition APIs


Take a look at the Paper and Project Page. Don’t forget to affix our 25k+ ML SubRedditDiscord Channel, and Email Newsletter, where we share the most recent AI research news, cool AI projects, and more. If you’ve gotten any questions regarding the above article or if we missed anything, be at liberty to email us at Asif@marktechpost.com

🚀 Check Out 100’s AI Tools in AI Tools Club


Ekrem Çetinkaya received his B.Sc. in 2018, and M.Sc. in 2019 from Ozyegin University, Istanbul, Türkiye. He wrote his M.Sc. thesis about image denoising using deep convolutional networks. He received his Ph.D. degree in 2023 from the University of Klagenfurt, Austria, along with his dissertation titled “Video Coding Enhancements for HTTP Adaptive Streaming Using Machine Learning.” His research interests include deep learning, computer vision, video encoding, and multimedia networking.


🔥 StoryBird.ai just dropped some amazing features. Generate an illustrated story from a prompt. Test it out here. (Sponsored)

LEAVE A REPLY

Please enter your comment!
Please enter your name here