Pixel Value Graphical Password Scheme: Analysis on Time Complexity performance of Clustering Algorithm for Passpix Segmentation

Abstract

Passpix is a key element in pixel value access control, containing a pixel value extracted from a digital image that users input to authenticate their username. However, it is unclear whether cloud storage settings apply compression to prevent deficiencies that would alter the file's 8-bit attribution and pixel value, causing user authentication failure. This study aims to determine the fastest clustering algorithm for faulty Passpix similarity classification, using a dataset of 1,000 objects. The source code for the K-Means, ISODATA, and K-Harmonic Mean scripts was loaded into a clustering experiment prototype compiled as Clustering.exe. The results demonstrate that the number of clusters affects the time taken to complete the clustering process, with the 20-cluster setting taking longer than the 10-cluster setting. The K-Harmonic Mean algorithm was the fastest, while K-Means performed moderately and ISODATA was the slowest of the three clustering algorithms. The results also indicate that the number of iterations did not affect the time taken to complete the clustering process. These findings provide a basis for future studies to increase the number of clusters for better accuracy.