UFDR Agent: An AI-Driven Framework for Automated Analysis of Universal Forensic Extraction Device Reports in Digital Forensic Investigations

Universal Forensic Extraction Device Report (UFDR) files generated from seized smartphones and computers routinely contain gigabytes to multiple terabytes of chats, call logs, images, videos and metadata, and manually reviewing this volume of evidence can take investigators weeks per device. This paper presents UFDR Agent, an AI-based analysis platform designed to automate the ingestion, structuring and querying of UFDR data using a combination of natural-language processing, computer vision and vector-similarity search. The system parses heterogeneous UFDR exports into a canonical relational schema, applies large-language-model-based intent detection to translate an investigator’s questions into structured queries, and uses vision-language and face-detection models to caption, tag and embed image evidence (and, prospectively, video) for semantic retrieval. Entity-extraction and link-analysis modules surface cryptocurrency addresses, international phone numbers, URLs and suspicious keywords, and the platform is designed to produce structured, court-oriented summary reports with timelines and an audit trail. A working prototype ingests approximately 8,900 records in 30 seconds and answers structured queries in under one second on commodity, offline hardware. This paper describes the motivation, related work, system architecture, implementation and evaluation of the proposed platform, and discusses its feasibility, expected impact on investigation timelines, and directions for future development.