Abstract: With invocations of a software development pipeline, organization specific remediations/fixes for a software project can be learned from scanning results of code submissions (e.g., commits or merges) across an organization for a software project(s). Fixes of detected program code flaws can be detected and/or specified across scans and associated with flaw identifiers and used for training machine learning models to identify candidate fixes for detected flaws. This ongoing learning during development propagates fixes created or chosen by experts (e.g., software engineers working on the software project) relevant to the software project. The experts can choose from suggestions mined from the learned fixes of the organization and suggestions generated from a pipeline created with the trained machine learning models. The selections are then used for further training of the machine learning models that form the pipeline.

Abstract: To preserve privacy when leveraging organization-specific remediation knowledge for flaw remediation across organizations, program code is deidentified to remove code which potentially identifies its source/origin. Deidentification operates based on structure of flaws and fixes at the level of source code constructs based on an abstract syntax tree (AST) or other structural context representation of a fix and corresponding flaw. Potentially identifying portions of a fix indicated in its AST are determined and modified (e.g., removed or obfuscated) without impacting AST structure. Deidentified remediation knowledge originating from different organizations is used to train a fix suggestion model(s) which learns structural context of fixes and corresponding flaws and, once trained, generates predictions indicating suggested fixes to flaws based on structural contexts of the flaws.