Google has opened the source code for two privacy-enhancing technologies (PETs), including AI systems for blurring objects in video.
🎁 A gift for you: New ways to provide a safer online experience for everyone!
From an #opensource #ML tool that instantly blurs objects to enhanced data encryption, we’re committed to continue designing technology that protects your privacy → https://t.co/a00el0MWoA pic.twitter.com/eAzsKrlWrc
— Google Developers (@googledevs) December 22, 2022
Both tools are part of Google’s Protected Computing initiative.
“The project aims to transform the ways, times and locations of data processing to ensure their confidentiality and security,” the tech giant’s site says.
The Magritte video-blurring tool is already available on GitHub. The system uses machine learning to detect objects in videos and blur them as soon as they appear on screen.
Magritte can mask a variety of objects, including vehicle license plates and people’s tattoos.
According to the company, the tool will be of use to journalists seeking heightened guarantees of privacy.
“Using this open system, videographers will be able to save time blurring objects in videos, given the high accuracy of detection by the machine-learning algorithm,” Google’s blog says.
The Fully Homomorphic Encryption (FHE) Transpiler enables developers to perform computations on encrypted data without accessing user-identifying information.
The company said the system could help in government and financial sectors, as well as healthcare.
“FHE Transpiler will be useful where heightened guarantees of security in handling confidential data are of paramount importance,” Google adds.
According to company representatives, PET is becoming mainstream, given that they were previously seen as little more than “academic exercises”.
In June the White House заявил that these technologies would allow researchers and other professionals to gain insight into data privacy without access to the data themselves.
In December Google unveiled an AI algorithm for handwriting recognition of doctors.
In the same month, the tech giant created the Simple ML add-on for developing AI models in Sheets.
In November, Google taught a robot to program itself.
Subscribe to ForkLog News on Telegram: ForkLog AI — all the news from the world of AI!