Detection of spam webpages using a new improved H-index centrality and weighted H-index centrality

Web spam is an unethical tactic used to manipulate search engine rankings by exploiting algorithmic loopholes. Detecting such spam pages remains a significant challenge for search engines aiming to provide relevant and trustworthy results. Centrality measures, derived from graph theory, are commonly used in network analysis to identify the most influential nodes, and several metrics have been proposed for unweighted and undirected networks. In this article, we propose two improved methods based on H-index centrality for weighted web graphs, namely Improved H-index Centrality and Improved Weighted H-index Centrality. These methods incorporate both the strength of neighbouring webpages and the weight of connecting links, thereby evaluating not only the number of connections but also the influence propagated through them. The proposed measures are evaluated on two web graphs and compared with existing centrality measures for spam detection. The performance is analysed using the distinctness metric, which measures their ability to differentiate between spam and trustworthy pages. Additionally, a Gaussian Naïve Bayes classifier is employed to assess classification performance, where the proposed measures demonstrate higher accuracy in identifying spam webpages. Experiments conducted on the WEBSPAM-UK2006 dataset show that the proposed methods outperform traditional centrality measures in distinguishing spam from legitimate webpages.