Temporal fingerprints for identity matching across fully encrypted domains

Area of Science:

• Computer Science
• Network Science
• Cybersecurity

Background:

• Coordinated inauthentic behavior (CIB) poses significant threats to societal stability, markets, and security.
• Generative AI exacerbates these threats by enabling sophisticated content creation and amplifying malicious actors' influence.
• Detecting CIB is challenging due to cross-domain activities of pseudonymous profiles and privacy constraints limiting content analysis.

Purpose of the Study:

• To propose a robust and scalable cross-domain identity matching framework for detecting coordinated inauthentic behavior.
• To develop a method independent of content or interaction data, focusing on behavioral dynamics.
• To provide a theoretical and methodological contribution to combating digital threats.

Main Methods:

• Developed a cross-domain identity matching framework based on bursty dynamics.
• Evaluated the framework's performance against state-of-the-art temporal and structural approaches.
• Assessed scalability and resilience to incomplete data across multiple online marketplaces.

Main Results:

• The proposed framework outperforms existing methods in detecting coordinated inauthentic behavior.
• It demonstrates resilience to incomplete data, identifying 35% of profiles after 52 weeks.
• Achieved an Area Under the Curve (AUC) of 0.78 when matching identities across 500 marketplaces with over 250,000 daily traders.

Conclusions:

• The bursty dynamics-based framework offers a scalable and effective solution for cross-domain identity matching.
• This approach enhances the ability to combat coordinated inauthentic behavior in complex digital ecosystems.
• Framing identity matching within a "network of networks" perspective reveals propagation patterns of coordinated behavior.