Cognitive Contextual Framework for Cross Channel Phishing Detection
1P.Siva Sai Santhosh, 2 S Amrutha ,3 A Devi Sri,4 D.Yaswanth Siva Kamalakar,5K.Deva Manikanta, 6Dr. K.N.V.R Kumar
¹²³⁴⁵Department of Computer Science and Engineering, Lingaya's Institute of Management and Technology, Vijayawada, India
⁶Ph.D, Professor, Department of Computer Science and Engineering, Lingaya's Institute of Management and Technology, Vijayawada, India
Abstract
Phishing schemes have evolved to be more advanced, using psychological manipulation, cross-channel delivery systems (ex: SMS, email, and URLs) and adversarial obfuscation to circumvent conventional detection methods. Current methods tend to be context-blind, do not combine multi-channel cues, and have little interpretability to end users. The paper suggests a Cognitive Contextual Framework of Cross-Channel Phishing Detection, which is a modular and scalable architecture that integrates machine learning, natural language processing, and cognitive analysis to improve the accuracy and explainability of detection. The model combines six complementary detection elements: (i) LightGBM to classify URL-based lexical features, (ii) XGBoost to classify SMS and email text, (iii) DistilBERT to classify deep semantic features, (iv) a cognitive vulnerability module to quantify social engineering cues, including urgency, fear, authority, and reward, (v) an online learning mechanism to detect drift, The system is implemented with FastAPI backend and React-based web and Kotlin-based mobile interfaces, and real-time inference and user feedback integration are supported. Empirical testing shows that the suggested framework can attain a detection accuracy of about 85 percent in a variety of communication channels, and is resistant to adversarial perturbations and linguistic diversity. Moreover, the system produces human-readable explanations to every prediction, enhancing transparency and user trust. The findings indicate the usefulness of using cognitive and contextual cues as the first-order features in phishing detection systems.
Keyword- Cross-channel phishing detection; cognitive security; social engineering detection; adversarial machine learning; DistilBERT; XGBoost; LightGBM; online learning; concept drift; explainable artificial intelligence (XAI); cybersecurity analytics; natural language processing.
