DOI: 10.53508/ijiam.1780395 ISSN: 2667-6990

CLASSIFICATION OF VISITORS AND IDENTIFYING INTRUDERS IN A NETWORK

Nada Hejazi, Enver Özdemir
Show PDF Cite
In today's digital ecosystem, accurately identifying user intentions on commercial websites is essential for enhancing security and optimizing user experience. This research introduces a phased classification approach for distinguishing website users into distinct groups based on behavioral patterns, addressing the nuanced challenge of distinguishing between genuine customers, casual browsers, and potential intruders. The proposed two-phase classification framework first differentiates customers from browsers, followed by a second phase that classifies browsers into either normal or suspicious users. This staged approach improves accuracy by focusing on progressively finer distinctions between user types, capturing subtle behavioral nuances that might otherwise be overlooked in a single-step model. Due to limitations of existing datasets, we generated a synthetic dataset that simulates realistic user behaviors, incorporating diverse features such as session duration, pages viewed, VPN usage, and geolocation consistency. A CatBoost classifier was employed in each phase for its ability to handle categorical data effectively. Experimental results demonstrate that this two-phase model achieves high accuracy, underscoring its potential for precise, context-aware user segmentation in online environments. This work offers a significant contribution to cybersecurity and user segmentation, advancing strategies for dynamically distinguishing user groups in digital spaces.

More from our Archive