Cyber ​​experts train Lensa-like apps, Telecom News, ET Telecom Lens

Chennai: Turning selfies into unique, colorful avatars using Lensa AI is all the rage these days, with celebrities like Deepika Padukone uploading their pictures using the photo editing app.

Within weeks of its launch, Lensa AI became the most downloaded free photo app in the US and the sixth most downloaded in India.

As photography apps like Lensa AI, owned by Prisma Labs, grow in popularity, cybersecurity experts are worried.

They point out the potential misuse of sensitive personal data such as facial recognition and iris scanning, and warn users to be extra careful when uploading their photos to any such apps, which have recently come under increased scrutiny for their privacy policies and security features. .

Lensa AI is not alone. Other photo apps like NewProfilePicture and FaceApp have also come under scrutiny for their policies.

“I’m concerned about what they (Lensa AI) do with the images you give the app, the metadata within the images, and how they are used in the model that the AI ​​uses to generate the images,” Dominic DiFranzo, assistant professor at Lehigh University in Pennsylvania professor of its department of computer science and engineering told ET.

Privacy Policy

According to DiFranzo, Lensa AI is an artificial intelligence (AI) solution that builds new images from a large corpus of other images.

“Your images will be part of this corpus and will affect the type of images generated for other users. This doesn’t mean other users can use/see your photos, but the AI ​​will be trained on your images in some cases. very small way ” – says.

Cybersecurity experts warn users to read a company’s privacy policy before agreeing to terms of service, as they may unwittingly agree to share personal information. A strong password, biometric features or two-factor authentication can also help prevent abuse, they say.

According to Lensa AI, User Content is used solely for the operation or development of the application.

The company claims that its rights include the “reproduction, modification, distribution, creation of derivative works” of user content without any further compensation. User Content is defined not only as user-uploaded photos, but also any artificial intelligence-generated content that a user creates using the Application, such as an avatar.

According to its previous policy, uploaded photos or videos could also be used to make its affiliate algorithms and products perform better, but the company clarified in its latest update that no personal data is used to train Prisma’s other AI products. Labs.

Lensa AI also clarified that if a user uploads a photo or video depicting a friend or anyone else, it can only do so with their consent.

“If you upload someone else’s content to Lensa and we receive claims, you will indemnify Lensa for such claims (meaning we will seek financial compensation for those claims),” according to the terms of the contract.

Earlier this year, photo app NewProfilePicture went viral for allowing users to turn their images into illustrated portraits with an “AI-powered” update similar to Lensa AI. However, shortly after it took the internet by storm, news emerged that the service, which uses facial recognition technology, was sending users’ photos and data to Russia.

Another Russian-developed app, FaceApp, was also extremely popular in India in 2019, with actors and other celebrities wondering how the app would help them see their transformed selves as they age.

AI training

According to Kevin Curran, professor of cyber security at the University of Ulster, the most important threat to user privacy is if Lensa AI or any other artificial intelligence photo app does not adhere to their policy of only using images to train their AI algorithms.

If proper authentication isn’t used for such services, hackers can harvest all stored images, he says, adding that standard encryption techniques are essential to protect user images.

“Ultimately, you have to trust that the company will honor its terms. Of course, Lensa AI claims that it does not store any images, but simply processes them for users and uses the information from each image to train its neural network, which is the system its heart. AI,” says Curran.

When companies share details with a third party, it gives them a lot of leeway to profit from personally identifiable information, he says.

Lensa claims to collect other data, including information about a user’s mobile device and internet connection, including their IP address, device unique device identifier, operating system and mobile network.

But it made it clear that it does not give, share, sell or otherwise transfer users’ photos, videos or avatars to ad platforms, analytics providers, data brokers and information resellers.