Google recently made its Google Magika technology open source. The technology uses AI and helps security specialists accurately identify files.
Google Magika technology works with a self-developed, highly optimized deep learning model, the tech giant indicates. This model is trained using the so-called Keras framework or API. Keras ensures high-speed debugging of code, better elegance and awareness of code, and better maintenance and deployability. This is through XLA compilation with JAX and TensorFlow.
In addition, AI or ML models developed with Keras are faster to deploy to different environments (server, mobile, or browser-based). This is because of built-in components from the TensorFlow and PyTorch ecosystems.
Google Magika is further only 1 MB in size and uses Open Neural Network Exchange (ONNX) as an inference engine. This ensures that files are identified in milliseconds.
Faster performance
The tech giant mainly emphasizes the speed of the tool, as this is the biggest advantage Magika provides. The tool is said to outperform conventional file identification with a more than 30 percent higher accuracy boost. Also, the tool would provide up to 95 percent greater accuracy for traditionally more difficult and also potentially dangerous VBA, JavaScript and Poweshell files.
Magika for better security
Google uses the technology to provide more security for users of Google services. To do this, Magika technology routes Gmail, Drive and Safe Browsing files to the appropriate security and content policy scanners.
Google improves file identification by 50 percent through Magika. This is in comparison with a previously used file identification system based on handwritten rules. As a result, Google says it is able to scan 11 percent more files, bringing the number of unidentified files down to 3 percent.
Previous initiatives
Google has been implementing AI to scan for malicious content in its solutions and applications for some time. Back in November last year, Resilient and Efficient Text Vectorizer (RETVec) technology was introduced. This so-called multilingual text processing model allows the tech giant to detect malicious content in Gmail, including spam and phishing messages.
Google Magika is now available via GitHub.
Also read: Gmail gets biggest security update in years
16 hours ago
