- Tram Ho
Microsoft has revealed how the software giant can fix more than 30,000 bugs generated by 47,000 programmers each month. Microsoft Chief Security Officer Scott Christiansen said Microsoft is tracking bugs on the GitHub and AzureDevOps repositories, but there are too many bugs that can’t be used with traditional methods.
Microsoft had to use up to 20 years of data history, including 13 million jobs and errors to create a “machine” machine-learning that can identify security flaws with accuracy up to 99%. With this tool, Microsoft engineers can find serious security errors that need to be fixed immediately.
“Our aim is to create a machine-learning system that can find and divide errors into serious or non-serious, confidential or non-secure, with a level of accuracy close to that of security experts. top secret, ” said Christiansen.
Not only do Microsoft’s systems automatically detect errors, they can be classified into extremely dangerous, serious, or unaffected forms. From there, Microsoft engineers will prioritize focusing on solving the most dangerous errors first.
Security and data science experts have worked together at Microsoft to build this system. Machine-learning systems are also constantly being trained with new data, to ensure that it is not obsolete.
With this system, Microsoft can find errors with huge volumes with accuracy up to 99%, and accurate classification of up to 97%.
Source : Genk