CrowdStrike's research focuses on enhancing cybersecurity through machine learning techniques such as MOLD (Modeling Label Dissonance), which aims to address label noise in data used for malware detection. Label noise, originating from incorrect or imperfectly labeled data, poses challenges to machine learning models in cybersecurity settings, where precise labeling is complex and expensive. MOLD helps identify and correct mislabeled data, improving model accuracy and reducing false positives by enhancing the training dataset with high-value samples. The approach encompasses both local and global noise modeling techniques, with MOLD offering a lightweight solution for pre-screening new samples. By incorporating Shapley values for data evaluation and employing ensemble models to mitigate overfitting, CrowdStrike demonstrates how MOLD can optimize training processes and boost model performance, particularly in static malware detection. The company plans to integrate MOLD into its training utility pipelines, pushing the boundaries of its malware detection models.