•  
  •  
 

Abstract

Cybersecurity in biometric systems is an urgent requirement due to their increasing use in identity verification, especially in smartphones, surveillance systems, and electronic transactions. These systems depend on distinctive and immutable biological features, such as fingerprints, Knuckles and facial features, making them potential targets for cyberattacks. In this context, the need to develop advanced security mechanisms, including encryption, forgery detection, and multi-factor authentication, has emerged to guarantee the confidentiality of biometric data and protect it from identity theft or manipulation. This trend emphasizes the need to integrate cybersecurity and biometric technologies to secure and ensure the reliability of systems in modern digital environments. Therefore, this research paper proposes an enhanced cybersecurity for a finger knuckle print recognition system (ECS-FKPRS) using an enhanced Rubik’s cube technique within the encryption algorithm. This ECS-FKPRS begins by extracting features of key components of the dorsal finger knuckle prints, namely the base-knuckle, the major-knuckle, the minor-knuckle, and the fingernails, using two deep learning models: ResNet-50 and MobileNet-V2. To complement this, the ECS-FKPRS uses a new segmentation method based on the Hands Landmark Module (MediaPipe Module) to detect components. After that, the resulting feature vectors are extracted for each model, which will be used in conjunction with the Rubik’s cube technique to generate an efficient and strong secret key. Then, the resulting knuckle features are encrypted with the Rabbit lightweight encryption algorithm. Furthermore, the Rubik’s Cube technique proposes the integration of logistic scrambling, permutations, and SHA-256 hashing with the encrypted biometric data, which constitutes an enhanced key generation method to ensure confidentiality and integrity against statistical attacks. The system achieved excellent results with security metrics showing outstanding results: NPCR (99.61% and 99.57%), and UACI (30.22% and 27.82%), and resistance to statistical attacks on the 11,076K Hands and the ’Poly U HD’ datasets. The ECS-FKPRS results prove that deep learning models using biometric data with an encryption algorithm are efficient. Moreover, it has high durability and security.

Download

Share

COinS