A ROBUST STEGO-CRYPTOGRAPHIC ANTI-PHISHING FRAMEWORK

Abstract

Phishing attacks manipulate victims through impersonation, often mimicking brand logos and digital identities. Primitive anti-phishing solutions lack accuracy and protection, while modern solutions suffer from biased supervised learning models, constant labeled data requirements, and more recently, generative adversarial perturbations, threatening artificial-intelligence (AI) and logo detection reliability.

The objective of this thesis is to design and develop a novel anti-phishing scheme using a stego-cryptographic approach. We present two novel eukaryotic immune system inspired algorithms: the Image Obfuscation Algorithm, which obfuscates images, and the Image Elucidation Algorithm, which elucidates them. Our framework further implements cryptography for sender verification, and intentionality for robust security.

Our zero-AI framework identifies legitimate senders with perfect accuracy and F-score, eliminating false positives and negatives to categorizes all 310,893 samples across 11 simulated conditions. Image quality metrics confirm our obfuscation and elucidation algorithm effectiveness. Additionally, original images restore identically for clients, laying a foundation for zero-day phishing detection.