MODERN POSSIBILITIES AND DIRECTIONS OF DEVELOPMENT OF EXPERT EXAMINATION OF DIGITAL AND SPEECH OBJECTS GENERATED BY NEURAL NETWORKS (DEEPFAKES)
Abstract and keywords
Abstract:
The use of modern technologies in cybercrime poses new challenges for the expert community. In particular, the rapid development of tools for generating synthetic audio, video, and text content (including deepfake technology) necessitates updating the scientific, methodological, hardware, and technical support for the examinations required to analyze such objects (phonoscopic, video technical, authorship examinations). Since 2019, the Forensic Science Center of the Ministry of Internal Affairs of Russia has been working in these areas. This article provides an overview of the measures taken and their results, and identifies current priorities. These include testing software designed to detect digital counterfeits, developing methodological approaches to using the obtained data in forming expert conclusions, and creating an interdepartmental expert methodology for the forensic examination of generated speech.

Keywords:
forensic analysis, deepfake, generated objects, neural networks, detec tor, forensic analysis methodology
Text
Text (RU) (PDF): Read Download
References

1. Shlyakhov A.R. Some problems and development paths of forensic technology, forensic examination, and legal cybernetics // Transactions of the Central Research Institute of Forensic Examinations. – Moscow, 1970. – Issue 2. – P. 29. (In Russ.).

2. Gromova A.V. Features of texts generated by neural networks in the context of forensic authorship // Theory and practice of forensic examination. – 2025. – No. 20 (4). – P. 50–58. (In Russ.).

3. Kazmin V.V. Features of the purpose of certain types of forensic examinations: Reference guide (interactive) / V.V. Kazmin et al. – Moscow: Forensic Center of the Ministry of Internal Affairs of Russia, 2025. (In Russ.).

4. Gromova A.V. Authorship examination of text messages: recommendations for a pre-expert assessment // Forensic examination. 2025. – № 2 (82). – P. 48–56.

Login or Create
* Forgot password?