Phishing the Phishers with SpecularNet: Hierarchical Graph Autoencoding for Reference-Free Web Phishing Detection

Dimensions overall score 8.0

GitHub Code Pulse

No public code linked for this paper yet.

• SpecularNet achieves competitive phishing detection performance with an F1 score of 93.9% on benchmark datasets.result 90%
• SpecularNet reduces inference time from several seconds to approximately 20 milliseconds per webpage compared to reference-based methods.technical 90%
• SpecularNet operates solely on domain name and HTML structure without relying on external knowledge bases or cloud services.method 90%
• SpecularNet uses a hierarchical graph autoencoding architecture with directional, level-wise message passing to model DOM structure.method 90%
• SpecularNet maintains resilience and accuracy against adversarial HTML manipulations.result 80%
• SpecularNet's reliance on DOM structure means it might miss phishing attempts that heavily manipulate these structures.limitation 85%
• SpecularNet enables fast, end-to-end inference on standard CPUs without requiring specialized hardware.technical 85%
• SpecularNet was evaluated against 13 state-of-the-art phishing detectors and showed competitive performance.result 90%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card