Synthetic Media Forensics and Directions for Development (The Case of Kazakhstan)

Abstract

This article reviews the new information challenges arising from the rapid development of generative artificial intelligence (AI) technologies and the modern forensic methods used to detect them. The current evolution of machine learning algorithms has begun to reduce the distinction between authentic content and synthetic media to a level imperceptible to the human eye. Specifically, the article analyzes the malicious potential of deepfakes in financial fraud schemes—such as bypassing banks’ KYC (Know Your Customer) protocols—and in political manipulations. The research also examines the experience of applying international benchmarks like FaceForensics++, exploring advanced verification approaches ranging from spatial analysis using convolutional neural networks (CNNs) and frequency analytics based on the discrete cosine transform (DCT), to physiological biometrics (rPPG) and innovative methods for evaluating audiovisual phonetic discrepancies. The paper investigates the proliferation of synthetic media in Kazakhstan’s information space, phonetic artifacts emerging in a bilingual Kazakh-Russian environment (temporal lip jitter in the articulation of uvular and velar consonants), and legal gaps in the new Law “On Artificial Intelligence”. In conclusion, the study substantiates the strategic necessity of transitioning away from the passive importation of foreign detectors toward creating localized national datasets, establishing an independent institute for algorithmic media audit, and integrating modern forensic tools into the fact-checking process.