Multimodal Data Fusion and Behavioral Analysis Tooling for Exploring Trust, Trust-propensity, and Phishing Victimization in Online Environments

摘要

Online environments, including email and social media platforms, are continuously threatened by malicious content designed by attackers to install malware on unsuspecting users and/or phish them into revealing sensitive data about themselves. Often slipping past technical mitigations (e.g. spam filters), attacks target the human element and seek to elicit trust as a means of achieving their nefarious ends. Victimized end-users lack the discernment, visual acuity, training, and/or experience to correctly identify the nefarious antecedents of trust that should prompt suspicion. Existing literature has explored trust, trust-propensity, and victimization, but studies lack data capture richness, realism, and/or the ability to investigate active user interactions. This paper defines a data collection and fusion approach alongside new open-sourced behavioral analysis tooling that addresses all three factors to provide researchers with empirical, evidence-based, insights into active end-user trust behaviors. The approach is evaluated in terms of comparative analysis, run-time performance, and fused data accuracy.