Full Length ArticleOptimal feature selection-based biometric key management for identity management system: Emotion oriented facial biometric system☆
Introduction
An identity management system (IMS) ensures security and reliability of digital transactions over a network. Digital services are provided to authenticated users who have valid credentials. Authorized levels of interaction amongst users, identity provider and service providers are to be regulated. Central administration, user self-service, role based access control and integrated user management are essential requirements for identity management systems [1]. Identity management increases efficiency and security of access control while decreasing complexity, cost and repetitive tasks. Integration of biometric traits with identity management systems brings in several advantages over username/password or token based authentication systems. Researchers have evinced a lot of interest in identifying novel physiological and behavioural traits for improving efficiency and security of identity management. Multi-modal biometrics is an area of active research to achieve increased efficiency. A significant challenge faced by researchers is in making biometric credential systems reliable and reproducible without sacrificing efficiency and efficacy of detection.
Biometrics has been used in broad range of applications such as e-commerce, physical and electronic access control, background verification and digital rights management [2]. Fusion of crytpographic keys with user specific traits addresses concerns of template security [3]. User specific traits are stored in smart card and utilized for matching features taken as input during validation [4]. Smart cards enable authentication with matching score within defined threshold value and are preferred over Session Initiation Protocol (SIP) servers [5]. The servers ignore user specific traits and depend on smart card for user verification [6], [7].
Conventional user name password or token based authentication systems have limitations. The major issue is that users are required to memorize usernames, passwords or maintain them securely making them susceptible to multiple attacks including dictionary attacks [8]. The solution to ensure secure storage of data involves substitution with cryptographic key [9], [10], [11]. Cryptographic models with smaller keys are easily broken while larger and complicated keys are difficult to remember. Larger keys are required to be stored securely and that in turn leads to threat vulnerability. Distribution of secret key is a challenging proposition in symmetric key cryptography [12], [13], [14].
Combination of biometrics and cryptography is a major area of research interest. Cryptographic keys are required to be precise to the last bit while user traits inherently show fuzzy nature due to intra-user variations. Biometric key release, key generation and key binding are approaches for combining cryptographic keys with user features [15]. Researchers are exploring newer traits with intelligent deep learning algorithms to perform user authentication. Keystroke dynamics and pattern of typing on mobile phones have been analysed with convolutional neural network to identify the user [16]. Deep learning techniques are employed for face recognition on mobile devices and for recognition systems based on finger veins [17], [18].
The major contribution of this work is:This work aims to develop a novel biometric key management system by considering facial emotions and key binding techniques. An environment with enhanced security is built by considering both facial and key features during encryption and decryption. Facial emotion features acquired by extracting Scale Invariant Feature Transform(SIFT) features are computationally complex for machine learning. A novel metaheuristic algorithm, namely, Fitness Sorted Deer Hunting Optimization Algorithm with Rooster Update (FDHOA-RU) has been developed for selecting optimal features from SIFT keypoints. The proposed FDHOA-RU ensures selection of optimal features, thereby, providing improved fitness resulting in higher efficiency during training. FDHOA-RU algorithm has been developed by integrating properties of Deer Hunting Optimization Algorithm (DHOA) and Chicken Swarm Optimization (CSO) algorithms. The FDHOA-RU algorithm circumvents problem of premature convergence by including updates from CSO algorithm. Key extraction from an input image is carried out by Double Random Phase Encoding (DRPE), Bose–Chaudhuri–Hocquenghem (BCH) encoding, shuffling and RSA encryption. The proposed approach has made key generation more specific and reduced the probability of information stealing. Facial emotions have been considered for neural network based training during encryption and decryption process along with the key, thus providing additional support to the biometric recognition system.
The paper is structured in following manner: An overview of recognition of facial emotions and challenges in binding cryptographic keys with user traits is provided in Section 2 with state-of-the-art works in key agreement protocol. Section 3 describes proposed key management for IMS with novel key binding procedure. Key encryption process is explained in Section 4. Selection of optimal features from facial emotions and training of neural network is described in Section 5. The process of image encryption and decryption with binding of user specific key is covered in Section 6. Section 7 evaluates the approach adopted and compares results obtained with other state of art metaheuristics algorithms. Finally, conclusions are drawn in Section 8.
Section snippets
Facial emotions
Facial emotions introduce natural aspects to conventional face recognition algorithms. Emotion feature set add additional dimension for enhanced versatility to detection accuracy [27]. Local Binary Pattern (LBP), Gabor filter, LBP with Three Orthogonal Planes, Graphics-processing based Active Shape Model (GASM) and Scale Invariant Feature Transform (SIFT) are some feature extraction methods explored to extract emotion feature set [28], [29], [30]. Research has been carried out in recent past
Proposed architecture
Multiple research have been carried out for binding cryptographic key with biometrics. Drawbacks of implemented methodologies of earlier studies include failure in secure transfer of encryption key and difficulty for user to store key in a secure manner. Diagrammatic representation of proposed key management is shown in Fig. 1, depicting both encryption and decryption process. The methodology introduces facial emotions based key binding system to encrypt an image with seven different emotions
BCH encoding and key shuffling
BCH coder encodes the encryption key. There exists a -error correcting code with parameters , , , for any positive integer . The linear cyclic code has ability to correct upto random across bit positions. BCH encoded key is split into blocks to perform shuffling operation. Each block of encryption key is designated by a number and the blocks are shuffled by user specific shuffling key . The shuffled key is indicated by .
SIFT-based feature extraction from facial emotions
SIFT identifies salient and stable feature points used for proposed key management. SIFT is image scaling and rotation invariant and moderately invariant to alterations in illumination allowing for alteration in occurrence of occlusion, noise, or clutter. SIFT algorithm is classified into four phases: Scale Space Extrema Detection, Key point Localization, Orientation Assignment and Key point Descriptor. The four phases are described briefly.
- (a)
Scale Space Extrema Detection: SIFT framework
Solution encoding and objective model
The proposed FDHOA-RU algorithm is used for both feature selection as well as training of neural network. Once SIFT features are given, optimal SIFT features are extracted. The weight function of neural network is optimized by FDHOA-RU for more accurate detection of authorized user. Encoding solution for optimal feature selection and classification is shown in Fig. 2.
Optimally generated features are represented as , in which and are
Experimental setup
The optimal feature selection-based biometric key management for IMS using facial emotions has been implemented in MATLAB 2018a installed on a PC with Windows 10 OS, 8 GB RAM and 64-bit operating system. Hardware platform for proposed work has not been implemented and has been left as future work. Facial biometric related to seven different emotions — normal, smile, sad, surprise, anger, fear, and disgust has been used. Following datasets have been used for experiments.
- (a)
Japanese Female Facial
Conclusion
A new IMS protocol has been demonstrated involving key extraction for encryption, feature extraction of different emotions, optimal feature selection, bit stream generation, and decryption process. The image is initially subjected to DRPE, wherein the key is created using CRPM and EFRT. Binary conversion, BCH encoding, shuffling and RSA encryption are performed on generated key, followed by extraction of features with different emotions of user facial biometric and optimal feature selection.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
References (43)
- et al.
An effective biometric cryptosystem combining fingerprints with error correction codes
Expert Syst. Appl.
(2012) - et al.
Biometrics based key management of double random phase encoding scheme using error control codes
Opt. Lasers Eng.
(2013) - et al.
A novel approach to fingerprint biometric-based cryptographic key generation and its applications to storage security
Comput. Electr. Eng.
(2018) - et al.
Generating stable biometric keys for flexible cloud computing authentication using finger vein
Inform. Sci.
(2018) - et al.
Anonymous biometrics-based authentication scheme with key distribution for mobile multi-server environment
Future Gener. Comput. Syst.
(2018) - et al.
Biometric cryptosystems: A new biometric key binding and its implementation for fingerprint minutiae-based representation
Pattern Recognit.
(2016) - et al.
A survey on new generation metaheuristic algorithms
Comput. Ind. Eng.
(2019) - et al.
A survey on optimization metaheuristics
Inf. Sci.
(2013) - et al.
Grey wolf optimizer
Adv. Eng. Softw.
(2014) - et al.
The whale optimization algorithm
Adv. Eng. Softw.
(2016)
A survey on identity management for the future network
IEEE Commun. Surv. Tutor.
Biometric-enabled authentication machines: A survey of open-set real-world applications
IEEE Trans. Hum.-Mach. Syst.
A security-enhanced alignment-free fuzzy vault-based fingerprint cryptosystem using pair-polar minutiae structures
IEEE Trans. Inf. Forensics Secur.
Biometric authentication: System security and user privacy
Computer
A biometric key generation method based on semisupervised data clustering
IEEE Trans. Syst. Man Cybern.
A new biocryptosystem-oriented security analysis framework and implementation of multibiometric cryptosystems based on decision level fusion
IEEE Trans. Inf. Forensics Secur.
Fingerprint combination for privacy protection
IEEE Trans. Inf. Forensics Secur.
A novel key generation scheme for biometric cryptosystems using fingerprint minutiae
Multibiometric cryptosystems based on feature-level fusion
IEEE Trans. Inf. Forensics Secur.
Performance and security-enhanced fuzzy vault scheme based on ridge features for distorted fingerprints
IET Biom.
Multiple-image encryption mechanism based on ghost imaging and public key cryptography
IEEE Photonics J.
Cited by (1)
Multimodal Biometric Authentication by Slap Swarm-Based Score Level Fusion
2022, Lecture Notes on Data Engineering and Communications Technologies
- ☆
This paper has been recommended for acceptance by Zicheng Liu.