Luca Debiasi

Impact of sensor ageing on iris recognition

Similar to the impact of ageing on human beings, digital image sensors develop ageing effects over time. Since these imagers ageing effects (commonly denoted as pixel defects) leave marks in the captured images, it is not clear whether this affects the accuracy of iris recognition systems. This paper proposes a method to investigate the influence of sensor ageing on iris recognition by simulative ageing of an iris test database. A pixel model is introduced and an ageing algorithm is discussed to create the test database. To establish practical relevance, the simulation parameters are estimated from the observed ageing effects of a real iris scanner over the timespan of 4 years.

Generation of iris sensor PRNU fingerprints from uncorrelated data

The photo response non-uniformity (PRNU) of a sensor can be used for various forensic tasks, such as source device identification, source device linking, classification of images taken by unknown cameras, integrity verification, authentication. To ensure good results a high quality PRNU fingerprint of the sensor is needed. This can be achieved by acquiring images with uncorrelated content and high saturation, which are then used to calculate the fingerprint. Generating the desired data with iris sensors is not trivial, since they mostly have limited configuration options. These limitations come either by the sensor itself or by the software used to acquire the data. We describe how the desired images can be acquired with different iris sensors and illustrate the challenges and problems faced during the acquisition pro-cess. Finally the impact of the PRNU fingerprints calculated from the uncorrelated data on the device identification results is evaluated in respect to the usage of correlated data.

Comparison of PRNU enhancement techniques to generate PRNU fingerprints for biometric source sensor attribution

Identifying the source camera which acquired a given image using the cameras PRNU is a well established task in image forensics, known as camera or device identification. Since digital image sensors are widely used to acquire biometric data, it is eligible that this task can also be performed with biometric sensors and the respective data. This has already been studied in literature. In this paper we focus on a slightly different task, which consists in clustering images acquired with the same sensor in a data set possibly containing images from an unknown number of biometric sensors. Previous work showed unclear results that have been difficult to interpret because of the low quality of the extracted PRNU. In this paper we compare the use of a PRNU enhancement technique to the use of special uncorrelated images acquired with known biometric sensors in this clustering context. We additionally propose extensions of existing source sensor attribution techniques using data from known sensors. Finally, the results of the enhancement approaches and the results using the uncorrelated data acquired with the known sensors are compared and an assessment on whether multiple sensor instances have been used in the different investigated data sets is given.

Blind biometric source sensor recognition using advanced PRNU fingerprints

Previous device identification studies on the iris sensors of the CASIA-Iris V4 database using PRNU fingerprints showed high variations regarding the differentiability of the sensors. These variations may have been caused by the usage of multiple sensors of the same model for the image acquisition. Since no speciic documentation on this exists we investigate the presence of multiple image sensors in the data sets. The images under investigation, furthermore, show a strong correlation regarding their content, therefore we make use of different PRNU enhancements approaches based on weighting the PRNU depending on the image content. The enhanced PRNU is used in conjunction with different forensic techniques to detect the presence of multiple sensors in the data sets. Finally, the results of the enhancement approaches and the results without any PRNU enhancement are compared and an assessment on whether multiple sensor instances have been used in the data sets is given.

Techniques for a forensic analysis of the CASIA-IRIS V4 database

The photo response non-uniformity (PRNU) of a sensor can be useful to enhance a biometric systems security by ensuring the authenticity and integrity of images acquired with a biometric sensor, e.g. by performing a source device identification. Previous studies regarding the feasibility of this application have been conducted on the CASIA-Iris V4 database by studying the differentiability of the sensors PRNU fingerprints. The results showed a high variation among the different subsets of the database. It was assumed that this high variation could either be caused by correlated data or that different sensors may have been used for the acquisition of the subsets. To investigate the latter case we perform a forensic investigation on the CASIA-Iris V4 database, since there is no specific documentation on the number of sensors used for the acquisition. We apply an existing forensic technique and we propose several novel forensic techniques to establish a ground truth of how many sensors have been used to a acquire a digital image data set in a blind manner and without any a priori knowledge.

Towards Drug Counterfeit Detection Using Package Paperboard Classification

Most approaches for product counterfeit detection are based on identification using some unique marks or properties implemented into each single product or its package. In this paper we investigate a classification approach involving existing packaging only in order to avoid higher production costs involved with marking each individual product. To detect counterfeit packages, images of the package’s interior showing the plain structure of the paperboard are captured. Using various texture features and SVM classification we are able to distinguish drug packages coming from different manufacturers and also forged packages with high accuracy while a distinction between single packages of the same manufacturer is not possible.

Real or Fake: Mobile Device Drug Packaging Authentication

Shortly, within the member states of the European Union a serialization-based anti-counterfeiting system for pharmaceutical products will be introduced. This system requires a third party enabling to track serialized and enrolled instances of each product from the manufacturer to the consumer. An alternative to serialization is authentication of a product by classifying it as being real or fake using intrinsic or extrinsic features of the product. Thereby, one approach is packaging material classification using images of the packaging textures. While the basic feasibility has been proven recently, it is not clear if such an authentication system works with images captured with mobile devices. Thus, in this work mobile drug packaging authentication is investigated. The experimental evaluation provides results on single- and cross-sensor scenarios. Results indicate the principal feasibility and acknowledge open issues for a mobile device drug packaging authentication system.

PRNU enhancement effects on biometric source sensor attribution

Identifying the source camera of a digital image using the photo response non-uniformity (PRNU) is known as camera identification. Since digital image sensors are widely used in biometrics, it is natural to perform this investigation with biometric sensors. In this study, the authors focus on a slightly different task, which consists in clustering images with the same source sensor in a data set possibly containing images from multiple unknown distinct biometric sensors. Previous work showed unclear results because of the low quality of the extracted PRNU. They adopt different PRNU enhancement techniques together with the generation of PRNU fingerprints from uncorrelated data in order to clarify the results. Thus they propose extensions of existing source sensor attribution techniques which make use of uncorrelated data from known sensors and apply them in conjunction with existing clustering techniques. All techniques are evaluated on simulated data sets containing images from multiple sensors. The effects of the different PRNU enhancement approaches on the clustering outcome are measured by considering the relation between cohesion and separation of the clusters. Finally, an assessment on whether the PRNU enhancement techniques have been able to improve the results is given.