Google LLC today introduced two open-source technologies designed to help companies more easily process user data in a manner that meets privacy requirements.
The first technology is a tool called Magritte for blurring objects such as license plates in videos. The other is a new version of FHE C++ Transpiler, a privacy tool that Google originally introduced last year. It allows applications to process encrypted datasets without decrypting them first.
Magritte is based on one of the search giant’s internal software projects. It uses artificial intelligence to automatically detect when an object containing sensitive data, such as a license plate, appears in a video. Magritte then blurs the object, removing the need for video editing teams to perform the task manually.
The AI features are powered in part by another open-source Google tool called MediaPipe. The latter tool enables developers to build AI applications that can run on devices with limited computing capacity, such as smartphones.
“This code is especially useful for video journalists who want to provide increased privacy assurances,” Miguel Guevara, a product manager at Google’s privacy and data protection office, detailed in a blog post. “By using this open-source code, videographers can save time in blurring objects from a video, while knowing that the underlying ML algorithm can perform detection across a video with high-accuracy.”
Google debuted Magritte today alongside a new version of FHE C++ Transpiler, an open-source tool it originally released last June. The tool makes it easier for developers to implement an encryption technology known as fully homomorphic encryption, or FHE. The technology has drawn significant interest from researchers in recent years because it could potentially be used to make enterprise applications more secure.
Enterprise applications store important data in an encrypted form to reduce the risk of cyberattacks. However, the data must be decrypted whenever it needs to be used. Decrypted files are more susceptible to cyberattacks because their contents can be easily accessed by hackers in the event of a breach.
The FHE encryption method on which Google’s FHE C++ Transpiler tool is based removes the need to decrypt data before it’s processed. As a result, the method allows companies to decrease the risk posed by cyberattacks.
In practice, however, using FHE to improve security is difficult because of several technical obstacles. One obstacle is that running FHE software currently requires a prohibitive amount of infrastructure. Another challenge is that the technology is difficult for developers to implement.
According to Google, its open-source FHE C++ Transpiler tool eases the task of deploying FHE. The tool can analyze a piece of code originally written to process decrypted data and automatically adapt it to run on data encrypted with FHE. As a result, developers can create applications capable of process encrypted data with less effort than the task would otherwise requires.
The new version of FHE C++ Transpiler that Google detailed today introduces several performance optimizations. The optimizations were implemented in the circuits that the tool uses to carry out processing. In computer science, the term circuit describes not an electronic component but rather a specialized series of computing operations carried out one after another.
Google’s engineers have reduced the size of the circuits FHE Transpiler uses to process data by 50%. The result, according to the search giant, is a significant performance improvement. The tool now requires less infrastructure to run and can carry out computations faster.
The significant amount of infrastructure currently necessary to run FHE software is among the main reasons the technology has not yet been widely adopted in the enterprise. By reducing hardware requirements, Google’s FHE tool could potentially make it easier for organizations to adopt FHE.
