Amplifying influence through coordinated behaviour in social networks

Area of Science:

• Social Network Analysis
• Computational Social Science
• Online Behavior Analysis

Background:

• Online malicious behaviors, including political misinformation, astroturfing, and organized trolling, pose significant threats to democratic systems.
• Existing research often focuses on identifying campaigns rather than the small, coordinated groups instigating them.
• Detecting these hidden networks is crucial for understanding and mitigating their real-world impact.

Purpose of the Study:

• To propose and validate a novel temporal window approach for revealing latent networks of cooperating accounts involved in online malicious behaviors.
• To develop a method capable of identifying coordinated amplification strategies and other inauthentic behaviors using only account interactions and metadata.
• To provide a framework suitable for near real-time application in detecting coordinated inauthentic behavior on social media.

Main Methods:

• A pipeline approach extracting social media post elements, inferring account connections based on coordination strategies, and building a weighted network.
• A novel community extraction method applied to the network to identify groups with high coordination evidence.
• A temporal windowing mechanism, sliding frames, and a decay factor to address the temporal dynamics of online interactions.

Main Results:

• The temporal window approach effectively detects coordinated groups and amplification strategies.
• The method demonstrates utility in identifying malicious online behaviors in politically relevant datasets.
• Validation against ground truth data using various analytical methods, including one-class classifiers, confirms the approach's effectiveness.

Conclusions:

• The proposed temporal window approach offers a robust method for uncovering hidden networks behind online malicious behaviors.
• This technique is valuable for understanding and combating coordinated inauthentic behavior that influences democratic processes.
• The approach's adaptability for near real-time analysis enhances its practical application in cybersecurity and social media monitoring.