Work

AI-Driven Discovery of Cybercrime Insights from Diverse Web Data

Year: 2026
Type: conference-paper
Abstract: The cybercrime is increasing in scale as well as intricacy and is abusing websites, emails, and online services in new manners that make the old defense mechanisms based on rules useless. This paper addresses the powerful need to extract actionable cyber intelligence on a highly heterogeneous web data. The goal is to create an integrated AI based system that can target phishing attacks with high accuracy on learning various, real world information sources. The suggested model unites three different types of data, including emails, URLs as well as HTML-based content as well as employs stringent preprocessing, metadata extraction advanced feature engineering along with SMOTE-based class balancing. It creates and optimizes a set of high-performance machine learning models, such as XGBoost, LightGBM as well as ExtraTrees. In order to achieve the highest predictive power a weighted soft voting ensemble is used whereby the system combines the complementary strengths of all models.The experimental results indicate that the ensemble is capable of attaining a majority accuracy of 82 which is better than all the individual classifiers and provides a very stable phishing detection with diverse attack patterns and feature modalities. These findings confirm the idea that multi-domain information integration alongside ensemble learning can significantly increase the power of cybercrime detection.This paper demonstrates that AI-based hybrid modeling is a scalable, intelligent and resilient solution to phishing detection.
Language: English
Open Access status: hybrid
OpenAlex ID: W7204189409
FWCI:
Cites: 10
Cited by:
Related to: 10