AMELIA - Automatic Machine Learning based Code Patching

Intelligent Automation, Incorporated (IAI) proposes an Automatic Machine Learning based Code Patching (AMELIA) framework to generate effective and efficient source code level patches. AMELIA first converts source code written in different programing languages into a language/target independent Intermediate Representation (IR). The collected IR code is then fed into Deep Neural Networks (DNNs) to train advanced Machine Learning (ML) models which will be used to detect and localize vulnerabilities from complex software systems. AMELIA also uses the DNN based models to represent a uniform knowledge on security vulnerabilities. AMELIA generates source code level patches, and will use existing analysis/optimization tools to optimize the placement of generated patches. Approved for Public Release | 19-MDA-10270 (18 Nov 19)