Fake Fingerprint Images Research Articles

A review on presentation attack detection system for fake fingerprint

Fingerprint recognition systems are susceptible to artificial spoof fingerprint attacks, like molds manufactured from polymer, gelatin or Play-Doh. Presentation attack is an open issue for fingerprint recognition systems. In a presentation attack, synthetic fingerprint which is reproduced from a real user is submitted for authentication. Different sensors are used to capture the live and fake fingerprint images. A liveness detection system has been designed to defeat different classes of spoof attacks by differentiating the features of live and fake fingerprint images. In the past few years, many hardware- and software-based approaches are suggested by researchers. However, the issues still remain challenging in terms of robustness, effectiveness and efficiency. In this paper, we explore all kinds of software-based solution to differentiate between real and fake fingerprints and present a comprehensive survey of efforts in the past to address this problem.

Integration of Statistical Method and Zernike Moment as Feature Extraction in Liveness Detection

Recently, fake fingerprints have been used to defeat fingerprint recognition systems. These fake fingerprints are created without the need for any expertise and use easily found materials. In this paper, a fake fingerprint detection method is proposed that employs a combination of eleven statistical methods and integrating them with Zernike Moment as the feature extractor. Based on the experimental results, the proposed method showed average classification accuracy, sensitivity and specificity of approximately 80% for all sensors used to capture fake fingerprint images fabricated by gelatine and latex materials.

An incremental learning method for spoof fingerprint detection

Spoof fingerprint detectors based on static features are built by learning a set of live and fake fingerprint images. These learning-based spoof detectors cannot accurately classify new or untrained types of fakes. To handle this problem, the existing spoof detector should be incrementally trained on the new types of fakes. This paper proposes a new spoof detection framework to learn new types of fakes incrementally without retraining the existing spoof detector repeatedly. The proposed model discriminates the newly learned fakes without serious loss of performance for the previously learned fakes and at the same time provides promising detection results for the various types of fakes. The proposed spoof detector integrates multiple “experts,” each of which shares the same structure but is separately trained for a different set of fake fingerprints. To detect a new type of fake fingerprint, a new expert exclusively trained on the new fake type is integrated into the spoof detector. Each expert consists of multiple support vector machines (SVMs) applied by an incremental learning algorithm (Learn++.NC), where each SVM adopts one of three texture features for spoof detection. Experimental results show the superiority of the proposed method compared with other methods in various scenarios.

Statistical anti-spoofing method for fingerprint recognition

Recently, attempts at spoofing biometric fingerprint recognition systems through fake fingerprints have been frequently reported. Most existing fake fingerprint detection methods require either additional sensors or complicated calculations. In this paper, a fake fingerprint detection method is proposed that employs combinations of six simple statistical moment features. These features are deviation, variance, skewness, kurtosis, hyperskewness, and hyperflatness of the fingerprints. In addition, the average brightness, standard deviation, and differential features are considered. Of all features, the best ones are selected in terms of the overlap ratio between real and fake fingerprint images. The multi-dimensional features are combined at the feature level through a support vector machine. Based on the experimental results, the proposed method showed classification accuracy of approximately 99%.

Fingerprint liveness detection using local texture features

The problem of fingerprint liveness detection has received an increasing attention in the last decade, as attested by the organisation of three editions of an international competition, named LivDet, dedicated to this challenge. LivDet editions and other works in the literature showed that the performance of current fingerprint liveness detection algorithms is not good enough to allow empowering a fingerprint verification system with a module aimed to distinguish alive from fake fingerprint images. However, recent developments have shown that texture-based features can provide promising solutions to this problem. In this study, a novel fingerprint liveness descriptor named binarised statistical image features (BSIFs) is adopted. Similarly to local binary pattern and local phase quantisation-based representations, BSIF encodes the local fingerprint texture into a feature vector by using a set of filters that, unlike other methods, are learnt from natural images. Extensive experiments with over 40,000 live and fake fingerprint images show that the authors’ proposed method outperforms most of the state-of-the-art algorithms, allowing a step ahead to the real integration of fingerprint liveness detectors into verification systems.

Classification of fingerprint images to real vs. spoof

Biometric identification is becoming a leading technology for identity management and security systems. Nonetheless, the use of counterfeit elastic fingerprints (|spoofing|) may break these measures. In this paper, we address the problem of fingerprint spoofing based solely on image features extracted from 2D fingerprint images. By combining several low-accuracy methods, a robust high-performance classifier for real vs. fake fingerprint images is constructed. Its high accuracy is demonstrated on a large fingerprint database. The method thus shows high potential for improving existing fingerprint authentication devices.