This document describes a proposed approach to rank cyber threat indicators of compromise (IOCs) matched in enterprise data based on reconstruction error from principal component analysis (PCA). The approach involves extracting IOCs from threat feeds, matching them to features engineered from proxy log and other data, decomposing the matched features with PCA, calculating a reconstruction error score to rank matches, and supplementing top ranked matches with contextual details to aid investigation. Future work may include adding an autoencoder model, incorporating analyst feedback, and using additional data sources.