A Segmentation-Free Approach For Iris-Based Authentication Using Blockchain: Preliminary Research Results

Abstract. Context: Authentication systems are integral to ensuring secure and reliable access to digital and physical infrastructures. Traditional biometric authentication methods, such as iris recognition, facial recognition, and finger- print recognition, each have their own strengths and limitations. Among these, iris recognition stands out for its high accuracy and low error rates, even in large-scale systems. Recent studies indicate that iris recognition has a false ac- ceptance rate (FAR) as low as 0.0001%, compared to fingerprint recognition, which can have a FAR of up to 0.1%, and facial recognition, which can reach up to 1% under similar conditions. Objective: This study presents a segmentation- free, end-to-end approach for iris-based authentication, leveraging deep con- volutional neural networks (CNNs) for feature extraction and classification. A proof of concept was conducted using the CASIA-Thousand-IRIS dataset to eval- uate the feasibility of the proposed method. Preliminary results show a testing accuracy of 92.5%, demonstrating the viability of the proposed approach. Con- clusions: The research presents the potential of applying blockchain to securely and decentrally manage identities, ensuring both enhanced accuracy and secu- rity for biometric authentication.