Causal Framework for Detecting Pathogenic Social Media Accounts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting pathogenic social media accounts rely on social media content, user profiles, and network structure, which require significant computational resources and are not always effective, especially when new types of malicious information are spread, and often lack access to necessary network structure information.
Innovation Solution
A causal framework that evaluates the likelihood of a social media account to cause a message to spread virally by analyzing historical cascade event data, using causality-based metrics to identify key users and potential pathogenic accounts without relying on content or network structure, thereby reducing computational demands and improving detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If content, user profile, and network structure information are collected and processed for detecting pathogenic accounts, then detection accuracy is improved, but computational resource consumption (CPU, memory, network bandwidth) increases significantly
Solution Approach 1:
The patent extracts and utilizes only the essential temporal sequence information from cascade events, removing unnecessary content, profile, and network structure data. This extraction approach maintains detection capability while significantly reducing computational resource requirements by processing only minimal necessary data.
Solution Approach 2:
Instead of starting with comprehensive data collection and then analyzing, the patent inverts the approach by directly observing temporal patterns in cascade sequences without requiring full content or network structure analysis. This inversion allows detection based on temporal dynamics alone, reducing computational burden.
2Measurement precision
If content information is used for detection, then detection capability is improved, but the system requires training of new models for previously unobserved topics, reducing adaptability
Solution Approach 1:
The patent creates a universal detection framework based on temporal cascade patterns that can detect different types of pathogenic accounts (terrorist supporters, water armies, fake news writers) across different topics without requiring topic-specific models. The temporal sequence analysis approach is universally applicable to any malicious information campaign.
Solution Approach 2:
The patent changes the detection parameter from content-based features to temporal sequence features of cascade events. This parameter transformation allows the system to detect various types of malicious accounts by analyzing the timing and sequence of events rather than the semantic content, enabling automatic adaptation to new topics.
3Measurement precision
If network structure information is collected for detection, then detection accuracy is improved, but the system cannot function when network structure information is not available (e.g., FACEBOOK API restrictions)
Solution Approach 1:
The patent extracts detection capability from network structure dependencies by focusing solely on temporal cascade event sequences. This extraction removes the requirement for network structure information, allowing the system to operate independently of platform-specific API restrictions and network data availability.
4Measurement precision
If comprehensive data collection is performed for early detection of pathogenic accounts, then detection thoroughness is improved, but the complexity of the detection system increases
Solution Approach 1:
The patent segments the detection process into simple temporal event sequence analysis, breaking down the complex task of pathogenic account detection into manageable temporal patterns. This segmentation simplifies the system architecture while maintaining thorough detection capability through systematic temporal pattern recognition.
Data Source
AI summary
Embodiments of a system and methods for detecting social media designed to spread malicious information to “viral” proportions are disclosed. Historical cascade event data from preselected social media accounts as well as information from related accounts is applied to one or more causality metrics to generate a set of causality values. Causality values are further refined and analyzed to determine how casual a user is with respect to a cascade as opposed to other similar users.