ControlFlag 1.1 is now available . ControlFlag is a machine learning-based code detection tool launched by Intel. Through the combination of artificial intelligence and training of more than 1 billion lines of code, it can automatically scan the source code of the repository for errors. Intel has applied it to software and hardware . production.
ControlFlag is a self-supervised special pattern detection system that learns control structures (if, etc.) in high-level programming languages (such as C/C++) by mining classic patterns from open source repositories (on GitHub and other Git platforms), and then uses Learned classic structures to detect anomalies in the user's own code.
Version 1.0 of ControlFlag , released last November, provides full support for the C language. v1.1 adds full support for the PHP language, this version fully supports learning typical patterns (training) and detecting abnormal patterns (inference) in if statements in C and PHP programs. Provide support for:
- Download the GitHub repositories for the C and PHP languages, mine conditional expressions and use them to train ControlFlag
- A dataset containing pre-mined conditional expressions from the GitHub repository
- Supports detection of abnormal conditional expressions in the target repository
Additionally, this release fixes handling typos ( #42 , #45 ) and typos ( #43 ).
Update notes: https://github.com/IntelLabs/control-flag/releases/tag/v1.1