Clustering and Topic Modeling of Phishing Texts: A Multi-Visualization Approach
Himanshu Tiwari · 2025
Phishing attacks remain a pervasive threat in cybersecurity, exploiting human vulnerabilities through deceptive textual content. This paper presents a novel approach to analyzing phishing texts by combining sentence embeddings, clustering, and topic modeling, enhanced with multiple visualization techniques. Using the "ealvaradob/phishing-dataset" from Hugging Face, we cluster phishing texts into semantically similar groups and extract overarching topics, visualizing results with UMAP, t-SNE, PCA, and advanced BERTopic plots. Our methodology reveals distinct phishing strategies (e.g., urgency-based scams, prize offers) and provides actionable insights for detection systems. Experimental results demonstrate the efficacy of our multi-visualization framework, offering a comprehensive tool for understanding phishing text patterns.