A New Feature Level‐Based Multi‐Biometric Identification System to Enable Secure and Reliable Communication Systems for Next‐Generation Smart Cities

ABSTRACT The concept of smart cities, which aims to improve urban lifestyles through innovative technologies and policies, has gained significant momentum over the last few years. As we transition into the next‐generation smart city era, it is crucial to review some of the fundamental technologies driving this evolution. Among these, security‐related challenges necessitate the implementation of effective authentication systems. This paper introduces a novel feature‐based multi‐biometric identification system designed to provide secure and reliable communication frameworks for next‐generation smart cities. The use of multi‐modal biometric identification for twins has long been a fascinating and valuable area of research. Because of its high reliability and acceptance, this approach significantly advances the domain of biometric identification for twins. The unique characteristics of twins are determined by variations in features extracted during the multi‐modal biometric process. However, many of these extracted features are superfluous, expanding the search space and complicating the generalization process. Therefore, the primary challenge lies in isolating the most salient features capable of accurately identifying twins through multi‐modal biometrics. To address this challenge, this research proposes an improved twin multi‐modal biometric identification framework. The proposed method employs a fusion of Dis‐Eigen features with the SailFish optimization algorithm to enhance both accuracy and efficiency. In this study, unique biometric features of twins were extracted and simplified using the Dis‐Eigen process, after which the Sailfish algorithm was applied to optimize model parameters and improve identification accuracy. As a result, the intra‐class similarity error decreased, while the inter‐class similarity error increased, improving twin identification. Consequently, several classifier applications achieved an identification rate exceeding 93%.