The purpose of this research is to produce a system composed of a detection engine and user-interface capable of detecting 75+ statically linked-libraries within Windows x64 binary executables with greater than 80% accuracy in under 5 minutes time per binary. The system will achieve this by innovating on the work shown by State-of-the-Art graphical models and modifying the modeling process to focus on models and features with extreme scalability to production workloads. This will be done through ablation studies, dimensionality reduction, and focusing on linear and tree-based models. The reason for producing such a system is to 1) improve the ability for defending organizations to be able to better inventory their software supply chains and 2) to provide more rigorous quantifications to be used inside of the software acquisitions process. This system will have numerous commercial applications along the axis of these defensive use cases, delivering value through the prevention and mitigation of software supply chain vulnerabilities helping to mitigate breaches that cost an average of greater than $3 million in 2020 according to the Ponemon Institute.
