Google releases ``Magika'', a tool that accurately identifies file formats using the power of AI, as open source
Google's research team has developed and released open source
Magika: AI powered fast and efficient file type identification | Google Open Source Blog
https://opensource.googleblog.com/2024/02/magika-ai-powered-fast-and-efficient-file-type-identification.html
Magika
https://google.github.io/magika/
Google launches AI Cyber Defense Initiative to improve security infrastructure
https://blog.google/technology/safety-security/google-ai-cyber-defense-initiative/
Modern society cannot exist without the Internet and computers, and cybersecurity is more important than ever to protect users from malicious attackers. However, a major challenge in cybersecurity is that while attackers only need to find and exploit a single vulnerability, defenders must always deploy the best defenses across the digital domain. We have to keep going.'' Cyberspace has become an advantageous situation for attackers, and in most cases defenders can only mitigate the threats that already exist.
Phil Venables , Google Cloud's chief information security officer, and others argue that the power of AI can reverse this situation called the 'defender's dilemma.' “AI can augment the work of security professionals and defenders in detecting threats, analyzing malware, finding and remediating vulnerabilities, and responding to incidents,” Venables et al. Masu.
Therefore, Google's research team developed a new tool called 'Magika' that uses AI to quickly and efficiently identify file formats, and released it as open source.
GitHub - google/magika: Detect file content types with deep learning
https://github.com/google/magika/tree/main
Malicious attackers disguise files containing malware or malicious code as legitimate files, then send them as email attachments or trick you into downloading them. The first step in preventing attacks via dangerous files is to accurately identify the file type.
Until now, much of the software that helps computers identify file types has relied on custom human heuristics . However, manual approaches are time-consuming, error-prone, and not very reliable against malicious attackers.
Magika, on the other hand, uses highly optimized custom deep learning models to identify file formats with extremely high accuracy. The research team also explains that because the size of the model is kept to 1MB, it is possible to identify files in milliseconds even on the CPU.
The graph below shows benchmark results comparing the accuracy of Magika with tools such as ExifTool and File magik . You can see that Magika's performance outperforms other tools.
You can check the file formats that can be identified by Magica and their accuracy from
Related Posts:
