Talk · International Conference on Cybersecurity, Digital Forensics, and AI Applications · June 1, 2026
File Carving with Binary Fragment Alignment: An Adjacency Matrix Optimization Approach
Yerassyl Yermekov, Madi Shayakhmetov, Kakim Sagyndykov, Abilkair Imanberdi, Aina Tazhiyeva, Aliya Zhetpisbayeva
Abstract
File carving enables file recovery independently of file system metadata, yet conventional methods fail when files are fragmented across non-contiguous disk blocks. This paper presents an approach integrating signature-based carving with binary fragment alignment through adjacency matrix optimisation. Fragment interrelationships are modelled using a weighted adjacency matrix with edge weights derived from boundary continuity analysis, structural marker identification, and sequential pattern matching. An optimisation routine determines the most probable fragment ordering by maximising the product of adjacency scores. Experimental evaluation on datasets containing fragmented JPEG, PNG, and PDF files demonstrates recovery rates of 87.3% for bifragmented files and 71.8% for trifragmented files, with an overall F1 score of 85.8%, significantly outperforming conventional carving tools and remaining competitive with recent machine-learning-based reconstruction methods while preserving full interpretability. We additionally report results on real-world fragmented disk images and a quantitative runtime and memory scalability analysis. The adjacency matrix heatmap provides forensic investigators with interpretable insights into fragment relationships.
More by Yerassyl Yermekov
- NEW APPROACH OF BYTE-LEVEL RECONSTRUCTION FOR ENHANCED DIGITAL FORENSIC DATA RECOVERY BASED ON MLScientific Journal of Astana IT University · 2026
- FORENSIC RECOVERY OF VIDEO EVIDENCE FROM DAHUA DHFS4.1 SURVEILLANCE SYSTEMS AFTER FORMATTINGScientific Journal of Astana IT University · 2026
- NotPetya: Anatomy of a Cyberweapon Disguised as RansomwareInternational Conference on Cybersecurity, Digital Forensics, and AI Applications · 2026
- Forensic Data Extraction and Recovery from Unmanned Aerial Vehicles: A~Framework for DJI-Based PlatformsInternational Conference on Cybersecurity, Digital Forensics, and AI Applications · 2026
