Self-citation, while a natural part of scholarly communication, can be exploited to artificially inflate citation counts and, by extension, an author’s perceived impact. Accurate detection of self-citations is thus essential for maintaining academic integrity. However, distinguishing legitimate self-citations is challenged by issues of author name ambiguity, where multiple researchers may share identical or similar names. In this work, we propose a hybrid system for self-citation detection that leverages an ensemble of author name disambiguation techniques. Our approach integrates Random Forests, Gradient Boosted Trees, RankMatch, and Novel String Processing algorithms, along with redundant ID identification, to systematically identify self-citations with high confidence. We validate our method through extensive experiments on two datasets, demonstrating robust performance with precision, recall, and F1 scores close to or achieving 1. Our results suggest that applying name disambiguation techniques substantially improves the reliability of self-citation detection, thereby providing a more accurate evaluation of academic metrics.