This story was originally published on HackerNoon at: https://hackernoon.com/new-study-shows-random-forest-models-can-spot-80percent-of-vulnerabilities-before-code-merge.
Machine-learning framework using Random Forest achieves ~80% vulnerability recall and 98% precision in real-world code review and deployment scenarios.
Check more stories related to tech-stories at: https://hackernoon.com/c/tech-stories. You can also check exclusive content about #ml-security-framework, #aosp-security, #ml-classifier, #secure-code-review, #software-security-testing, #upstream-code-security, #ai-code-review, #android-security, and more.
This story was written by: @codereview. Learn more about this writer by checking @codereview's about page, and for more stories, please visit hackernoon.com.
The study evaluates a machine-learning framework for predicting vulnerable code changes, showing Random Forest delivers the highest accuracy, robust performance across reduced feature sets, and significantly stronger precision and recall during real-world online deployment using six years of AOSP data.