Forensic checks for insurance companies trying to verify signatures, can be extremely time consuming. Human resources often are overloaded with manual forensic checks, causing delayed clearances for every forensic check request. We used machine learning to study a set of forged and authentic signatures between the real world identity proof documents and insurance application documents submitted by applicants. We then created a few ML models to denoise and then compare submitted and questioning signatures to identify forgery. Our solution is now deployed by a leading MNC for their forensic checks.