Evaluating the Effectiveness of Cyber Cognitive Attacks: A Sentiment-Based Approach

Author Information

Bonnie Rushing, University of Colorado Colorado Springs, United States
William Hersch, United States Air Force Academy, United States
Kora Gwartney, University of Colorado Colorado Springs, United States
Shouhuai Xu, University of Colorado Colorado Springs, United States

Abstract

Cyber cognitive attacks are a growing threat, yet their effectiveness remains difficult to measure systematically. We introduce the Cyber Cognitive Attack Effects Chain—Resonance (affective engagement), Proliferation (spread), and Influence (impact)—and propose a lightweight, data-driven method linking sentiment, topic, and engagement metrics. Resonance is operationalized via headline sentiment, while proliferation is captured through engagement percentiles, enabling tail-aware comparisons of narrative spread. Applying this framework to the FakeNewsNet–PolitiFact dataset (1,056 articles), we find that false narratives are more emotionally polarized, skew more negative, and achieve substantially higher median engagement than real news, indicating broader and more consistent virality. Topic–sentiment interactions reveal high-risk pockets (notably false–health and false–celebrity) that amplify spread, while real political content dominates aggregate engagement through a small number of highly amplified items. These patterns align with DISARM execution tactics, providing quantitative support for Maximise Exposure (TA17) and Deliver Content (TA09). The framework generalizes across datasets and offers practical implications for detection and defense, including sentiment-first triage and topic-aware monitoring of high-variance narratives. We conclude by outlining limitations and directions for extending measurement to downstream influence.

Save as a Favourite