New survey proposes proactive lifecycle-based framework to detect GenAI-driven synthetic narratives

The proliferation of GenAI-generated synthetic content is overwhelming traditional reactive detection methods. A new survey by Chung et al., accepted for IEEE Access, argues for a paradigm shift toward proactive detection of emerging inauthentic narratives. The authors introduce a unified lifecycle-based taxonomy centered on the C5 Interaction Model, which breaks down adversarial campaigns into Context, Causes, Content, Cycle of Amplification, and Consequences. This structure integrates machine learning with social science to differentiate synthetic amplification patterns from authentic baseline traffic. The survey systematically reviews state-of-the-art techniques for modeling narrative creation, seeding, and propagation, including Coordinated Inauthentic Behavior (CIB) analysis, epidemiological models, and Hawkes processes.

Beyond identification, the paper catalogs proactive detection methods tailored to each stage of the C5 model: anomaly detection in high-dimensional embeddings, unsupervised coordination detection on multi-layer graphs, and agentic AI systems. It also addresses pressing challenges like rapidly evolving threats and multi-level distributional drift. The authors outline a future research agenda focused on detecting anomalous clusters and building anticipatory, resilient information ecosystems. For tech professionals, this survey is a comprehensive guide to moving from reactive defense to preemptive resilience against GenAI-powered disinformation campaigns.