Loading article...
Loading article...
Stream/Area:
-
Subject:
-
Paper ID:
IJSREM67750
Volume:
10 |
10 |
Oct 2026
Pages:
1-8
Publication Frequency:
Monthly (12 issues per year)
License:International Journal of Scientific Research in Engineering and Management follows the Creative Commons Attribution 4.0 (CC BY 4.0) License, allowing free use and distribution with proper attribution.
SOC- Cyber Security and Forensics
MIT ADT University
Cyber Security and Forensics
MIT ADT University
Cyber Security and Forensics
MIT ADT University
SOC- Cyber Security and Forensics
MIT ADT University
A deepfake verdict is only as dependable as the record that surrounds it. The first version of our Deepfake Forensics Tool stored a SHA-256 digest of every submitted file and a JSON audit trail, but both sat in the same database as the analysis results, so anyone able to write to that database could rewrite the history without leaving a trace. This paper closes that gap with a lightweight tamper-evident integrity layer for AI-based media forensics. Each custody event is appended to a per-evidence hash chain; the chain is committed to a Merkle tree; and, when the report is produced, a seal that binds the file hash, chain head, Merkle root and report hash is signed with Ed25519 by a service outside the application server, while the seal digest is written to an independent append-only anchor store. A verifier needs only the public key, the anchor record and the evidence file. We built the layer as a Python prototype and replayed 1,000 synthetic analysis sessions through nine attack scenarios (value edits, timestamp edits, deletion, insertion, reordering, tail truncation, full chain recomputation, report modification and seal substitution) against four configurations. The complete design rejected all 9,000 tampered cases and accepted all 1,000 untouched sessions, whereas a plain JSON trail missed four attack classes entirely and a fifth in part, and a hash chain without an external anchor missed tail truncation and chain recomputation. On a single virtual CPU, appending an entry took about 5 µs, sealing a ten-event session about 0.12 ms, and verifying a 100,000-entry log about half a second.
KEYWORDS
A detector's verdict becomes evidence only when the record around it can be trusted. We showed that the first version of our tool leaves that record editable by anyone who controls its database, and we proposed a layer that fixes this with four ingredients: a hash chain over custody events, a Merkle commitment, an Ed25519-signed seal that covers the report, and an external anchor. The attack campaign explains the role of each ingredient. The chain catches local edits, the signed seal catches truncation and recomputation, the report digest catches document changes, and the anchor catches substituted seals. All 9,000 tampered sessions were rejected and all 1,000 untouched sessions were accepted, at a cost of about 0.12 ms per session. The next steps are integration into the production tool, anchoring to an external time authority and evaluation with practising forensic examiners.
[1] R. Chesney and D. Citron, “Deep fakes: A looming challenge for privacy, democracy, and national security,” California Law Review, vol. 107, pp. 1753–1819, 2019. [2] A. Utikar, A. Pund, S. Shinde, and A. Bhondave, “Deepfake Forensics Tool: AI-driven multi-modal media authentication with forensic evidence reporting,” Int. J. Sci. Res. Eng. Manag. (IJSREM), vol. 10, no. 3, Mar. 2026, doi: 10.55041/IJSREM57455. [3] ISO/IEC 27037:2012, Information technology — Security techniques — Guidelines for identification, collection, acquisition and preservation of digital evidence, ISO, 2012. [4] Government of India, The Bharatiya Sakshya Adhiniyam, 2023 (Act No. 47 of 2023), s. 63 and the Schedule. [5] R. Tolosana, R. Vera-Rodriguez, J. Fierrez, A. Morales, and J. Ortega-Garcia, “Deepfakes and beyond: A survey of face manipulation and fake detection,” Information Fusion, vol. 64, pp. 131–148, 2020. [6] A. Rössler, D. Cozzolino, L. Verdoliva, C. Riess, J. Thies, and M. Nießner, “FaceForensics++: Learning to detect manipulated facial images,” in Proc. IEEE/CVF Int. Conf. Computer Vision (ICCV), 2019, pp. 1–11. [7] B. Dolhansky et al., “The DeepFake Detection Challenge (DFDC) dataset,” arXiv:2006.07397, 2020. [8] L. Verdoliva, “Media forensics and deepfakes: An overview,” IEEE J. Sel. Topics Signal Process., vol. 14, no. 5, pp. 910–932, Aug. 2020. [9] R. R. Selvaraju et al., “Grad-CAM: Visual explanations from deep networks via gradient-based localization,” in Proc. IEEE Int. Conf. Computer Vision (ICCV), 2017, pp. 618–626. [10] S. Haber and W. S. Stornetta, “How to time-stamp a digital document,” J. Cryptology, vol. 3, no. 2, pp. 99–111, 1991.
Publication Date
10/9/2026
Peer Review Type
Double-Blind Peer Review
Review Rounds: 2
Documents
Article Views
0
Total views since publication
PDF Downloads
7
Total downloads since publication
Geographic Reach
5
Digital Preservation: This article is permanently archived on Zenodo and complies with all data preservation standards across major indexing platforms.
Plagiarism Screening
This manuscript has been checked for plagiarism using TurnitinPlagiarismChecker software.
Conflict of Interest
The authors declare no conflict of interest related to this research work.
Funding Information
No funding information available for this research.
Related research in the same subject area
Loading similar articles...