Citation manipulation, encompassing practices such as citation cartels, coercive citations, and reference padding, poses a significant threat to the integrity of scholarly communication. This study introduces a data-driven framework for detecting and assessing anomalous citation behavior by leveraging reconstruction errors. The proposed approach utilizes Ant Colony Embedding (ACE), a biologically inspired algorithm that captures hierarchical graph clustering to analyze citation networks through controlled, sequential distortions. By iteratively masking and reconstructing citation links, the method quantifies link reliability based on reconstruction rates, where high reconstruction rates indicate credible citations and low rates suggest potential manipulation. ACE’s capacity to model global clustering structures and evaluate node relationships strengthens the framework’s ability to identify unreliable citations, thereby improving the reliability of academic assessments. Numerical experiments conducted on the well-known Cora dataset validate the method’s effectiveness in detecting suspicious citations, highlighting its potential applications in future stylometric and bibliometric analyses.

In this article we offer an algorithm recurrently divides a dataset by search of partitions via one dimensional subspace discovered by means of optimizing of a projected pursuit function. Aiming to assess the model order a resampling technique is employed. For each number of clusters, bounded by a predefined limit, samples from the projected data are drawn and clustered through the EM algorithm. Further, the basis cumulative histogram of the projected data is approximated by means of the GMM histograms constructed using the samples’ partitions. The saturation order of this approximation process, at what time the components’ amount increases, is recognized as the “true” components’ number. Afterward the whole data is clustered and the densest cluster is omitted. The process is repeated while waiting for the true number of clusters equals one. Numerical experiments demonstrate the high ability of the proposed method.