End-to-End Evaluation of an Integrated Machine Learning-Based Drone Media Recovery Toolkit

Abstract

With the growing use of Unmanned Aerial Vehicles (UAVs), commonly known as drones, in both civilian and military contexts, the ability to recover and analyze deleted drone media has become critical for forensic investigations and intelligence operations. This study introduces and evaluates an integrated machine learning-based toolkit designed to ensure accurate recovery, metadata validation, and geospatial analysis of tampered UAV datasets. Furthermore, it presents a comprehensive evaluation of an end-to-end ML-based forensic toolkit for recovering and analysing deleted drone media. The system integrates file carving, GPS prediction, metadata validation, and interactive geospatial visualization. This paper benchmarks each module, measures forensic accuracy, and reports on system-wide performance across diverse UAV datasets and tampered scenarios. The findings of the study demonstrate robust recovery accuracy, anomaly detection precision, and real-world readiness for forensic and intelligence use.