AGAVM Phase II

Seed Innovations designs and develops a prototype system, the Automatic Generation of Application Vulnerability Mitigations (AGAVM). AGAVM automatically generates source code level mitigations to potential vulnerabilities in application software based on weaknesses identified by commercial-off-the-shelf (COTS) source code analysis tools. AGAVM features a plug-in architecture to facilitate incorporation of source code analysis tools into the ‘identify weaknesses → recognize vulnerabilities → mitigate vulnerabilities’ process flow and is informed by a knowledge base of recognition and mitigation rules and coding solutions. AGAVM uses machine learning and other AI techniques to discover new rules and coding solutions to the knowledge base from external sources including the Common Weakness Enumeration (CWE) database, the Common Vulnerabilities and Exposures (CVE), static code analysis web sites, and source code repositories. For the prototype AGAVM, Seed Innovations applies modern techniques including inference, automatic code generation, machine leaning and AI methods for recognizing and mitigating software vulnerabilities at the source code level to help ensure the integrity of the military information infrastructure and help protect the warfighter against cyber threats. When deployed to meet MDA’s cybersecurity challenge, AGAVM enables automatic source code level mitigation of software vulnerabilities thereby substantially reducing the time required to validate software against vulnerabilities for inclusion in the information infrastructure. Approved for Public Release | 20-MDA-10643 (3 Dec 20)