Evaluating the Precision of a Cloud-Based Attendance System with Iris Recognition: Daugman's Algorithm Versus Convolutional Neural Network

A comparative study of Daugman's algorithm and Convolutional Neural Networks (CNNs) in a cloud-based biometric attendance system based on iris recognition is presented in this paper. The objective was to quantify their accuracy, reliability, and practicability for real-time deployment in intelligent systems. A collection of iris images was processed and analyzed using both algorithms. Group statistical analysis revealed that Daugman's method produced a better mean accuracy (97.68%) compared to CNNs (87.23%). Further, independent samples t-tests also confirmed the performance difference as statistically significant in favor of Daugman's method in terms of precise consistency. While Daugman's algorithm had an edge when it came to feature extraction accuracy, CNNs offered greater flexibility and scope for improvement through learning. Their adaptability to incorporation in the cloud also makes them appropriate for massive, dynamic attendance environments. Based on the research highlighted here, it is determined that Daugman's algorithm is most applicable in systems requiring high accuracy and low variance, whereas CNNs are better applied to large-scale, real-time AI-based attendance systems, particularly in multivariate and dynamic settings.