This document discusses a methodology for detecting vulnerabilities in software based on analysis of the project's Git history. It proposes an approach called HVD that considers whether lines of code were added or removed in code changes, which could improve precision over existing techniques. An evaluation using a dataset of over 350,000 commits found that HVD increased the area under the precision-recall curve by 18.8% compared to a baseline that ignores line additions and removals. Features related to computer resources like memory, CPU and networking were found to most significantly contribute to the classification model. The study demonstrates that automatically detecting vulnerabilities from Git data can produce results aligned with human intuition.