Andrey Makrushin

Automatic Generation and Detection of Visually Faultless Facial Morphs.

This paper introduces an approach to automatic generation of visually faultless facial morphs along with a proposal on how such morphs can be automatically detected. It is endeavored that the created morphs cannot be recognized as such with the naked eye and a reference automatic face recognition (AFR) system produces high similarity scores while matching a morph against faces of persons who participated in morphing. Automatic generation of morphs allows for creating abundant experimental data, which is essential (i) for evaluating the performance of AFR systems to reject morphs and (ii) for training forensic systems to detect morphs. Our first experiment shows that human performance to distinguish between morphed and genuine face images is close to random guessing. In our second experiment, the reference AFR system has verified 11.78% of morphs against any of genuine images at the decision threshold of 1% false acceptance rate. These results indicate that facial morphing is a serious threat to access control systems aided by AFR and establish the need for morph detection approaches. Our third experiment shows that the distribution of Benford features extracted from quantized DCT coefficients of JPEG-compressed morphs is substantially different from that of genuine images enabling the automatic detection of morphs.

Advanced techniques for latent fingerprint detection and validation using a CWL device

The technology-aided support of forensic experts while investigating crime scenes and collecting traces becomes a more and more important part in the domains of image acquisition and signal processing. The manual lifting of latent fingerprints using conventional methods like the use of carbon black powder is time-consuming and very limited in its scope of application. New technologies for a contact-less and non-invasive acquisition and automatic processing of latent fingerprints, promise the possibilities to inspect much more and larger surface areas and can significantly simplify and speed up the workflow. Furthermore, it allows multiple investigations of the same trace, subsequent chemical analysis of the residue left behind and the acquisition of latent fingerprints on sensitive surfaces without destroying the surface itself. In this work, a FRT MicroProf200 surface measurement device equipped with a chromatic white-light sensor CWL600 is used. The device provides a gray-scale intensity image and 3D-topography data simultaneously. While large area scans are time-consuming, the detection and localization of finger traces are done based on low-resolution scans. The localized areas are scanned again with higher resolution. Due to the broad variety of different surface characteristics the fingerprint pattern is often overlaid by the surface structure or texture. Thus, image processing and classification techniques are proposed for validation and visualization of ridge lines in high-resolution scans. Positively validated regions containing complete or sufficient partial fingerprints are passed on to forensic experts. The experiments are provided on a set of three surfaces with different reflection and texture characteristics, and fingerprints from ten different persons.

Visibility enhancement and validation of segmented latent fingerprints in crime scene forensics

Forensic investigators are permanently looking for novel technologies for fast and effective recovering of latent fingerprints at a crime scene. Traditionally, this work is done manually and therefore considered very time consuming. Highly skilled experts apply chemical reagents to improve visibility of traces and use digital cameras or adhesive tape to lift prints. Through an automation of the surface examination, larger areas can be investigated faster. This work amplifies the experimental study on capabilities of a chromatic white-light sensor (CWL) regarding the contact-less lifting of latent fingerprints from differently challenging substrates. The crucial advantage of a CWL sensor compared to taking digital photographs is the simultaneous acquisition of luminance and topography of the surface, extending the standard twodimensional image processing to the analysis of three-dimensional data. The paper focuses on the automatic validation of localized fingerprint regions. In contrast to statistical features from luminance data, previously used for localization, we propose the streakiness of a pattern as the basic feature indicating the fingerprint presence. Regions are analyzed for streakiness using both luminance and topography data. As a result, the human experts significantly save time by dealing with a limited number of approved fingerprints. The experiments show that the validation performance in terms of equal error rate does not exceed 6% even on very challenging substrates regarding high-quality fingerprints.

Handwriting biometrics: feature selection based improvements in authentication and hash generation accuracy

Biometric cryptosystems extend the user authentication functionality of usual biometric systems with the ability to generate robust stable values (also called biometric hashes) from variable biometric data. This work addresses a biometric hash algorithm applied to handwriting data and investigates the performance of both user authentication and hash generation scenarios. In order to improve the hash generation performance, some feature selection approaches are proposed. The intelligent reduction of features leads not only to a better ratio of collision/reproduction rates, but also improves equal error rates in user authentication scenario. Additionally, the parameterization of biometric hash algorithm is discussed. It has been shown that different quantization parameters as well as different features should be selected to achieve better performance rates in both scenarios. For the best semantic, symbol, the EER is improved from 8.30% to 5.27% and the CRR from 11.20% to 6.32%. Finally, the almost useful and needless features are figured out e.g. only 2 features are selected for every semantic in both scenarios and 10 features are never selected.

Benchmarking face morphing forgery detection: Application of stirtrace for impact simulation of different processing steps

We analyze StirTrace towards benchmarking face morphing forgeries and extending it by additional scaling functions for the face biometrics scenario. We benchmark a Benfords law based multi-compression-anomaly detection approach and acceptance rates of morphs for a face matcher to determine the impact of the processing on the quality of the forgeries. We use 2 different approaches for automatically creating 3940 images of morphed faces. Based on this data set, 86614 images are created using StirTrace. A manual selection of 183 high quality morphs is used to derive tendencies based on the subjective forgery quality. Our results show that the anomaly detection seems to be able to detect anomalies in the morphing regions, the multi-compression-anomaly detection performance after the processing can be differentiated into good (e.g. cropping), partially critical (e.g. rotation) and critical results (e.g. additive noise). The influence of the processing on the biometric matcher is marginal.

Forensic analysis: on the capability of optical sensors to visualize latent fingerprints on rubber gloves

Thin rubber gloves are worn by criminals to prevent depositing fingerprints at crime scenes and are favored because of their tight fit, allowing hands to remain dexterous. However, fingerprints may be recovered from the inside of the gloves. The high variety of glove materials does not allow for a unified forensic approach for gloves investigation. All approaches proposed so far imply intrusive destructive treatment of the evidence. In contrast, we investigate the applicability of two contactless non-destructive sensors, a chromatic white-light sensor (CWL) and a UV-VIS spectroscope (UVVS), for digitalizing latent fingerprints left on three rubber materials: vinyl, nitrile and latex. The sensors are used to explore the visibility of sebaceous fingerprints over time, with the focus on qualitative assessment of data acquisition scenarios. Experiments show that fingerprints on porous vinyl gloves become invisible to the naked eye within 15 minutes after deposition, making the substrate very challenging. Here, fresh fingerprints can be acquired only with CWL. Fingerprints on nitrile remain preserved between 2 hours and 2.5 days and can be better captured using UVVS due to the possibility of integrating images over a certain range of wavelengths. Fingerprints on non-porous latex remain almost unchanged for at least one month and can be successfully captured using either CWL or UVVS.

Modeling Attacks on Photo-ID Documents and Applying Media Forensics for the Detection of Facial Morphing

Since 2014, a novel approach to attack face image based person verification designated as face morphing attack has been actively discussed in the biometric and media forensics communities. Up until that point, modern travel documents were considered to be extremely hard to forge or to successfully manipulate. In the case of template-targeting attacks like facial morphing, the face verification process becomes vulnerable, making it a necessity to design protection mechanisms. In this paper, a new modeling approach for face morphing attacks is introduced. We start with a life-cycle model for photo-ID documents. We extend this model by an image editing history model, allowing for a precise description of attack realizations as a foundation for performing media forensics as well as training and testing scenarios for the attack detectors. On the basis of these modeling approaches, two different realizations of the face morphing attack as well as a forensic morphing detector are implemented and evaluated. The design of the feature space for the detector is based on the idea that the blending operation in the morphing pipeline causes the reduction of face details. To quantify this reduction, we adopt features implemented in the OpenCV image processing library, namely the number of SIFT, SURF, ORB, FAST and AGAST keypoints in the face region as well as the loss of edge-information with Canny and Sobel edge operators. Our morphing detector is trained with 2000 self-acquired authentic and 2000 morphed images captured with three camera types (Canon EOS 1200D, Nikon D 3300, Nikon Coolpix A100) and tested with authentic and morphed face images from a public database. Morphing detection accuracies of a decision tree classifier vary from 81.3% to 98% for different training and test scenarios.

Visibility Assessment of Latent Fingerprints on Challenging Substrates in Spectroscopic Scans

Our objectives for crime scene forensics are to find the substrates on which finger traces are visible in limited ranges of the electromagnetic spectrum using UV-VIS reflection spectroscopy and to determine the optimal ranges in the interval from 163 to 844 nm. We subjectively assess the visibility of fingerprints within detailed scans with a resolution of 500 ppi and compare the results with those of an automatic visibility assessment based on the streakiness score. Ten different substrates are evaluated, each with three fingerprints from different donors. Streakiness score is confirmed to be a suitable fingerprint visibility indicator on non-structured substrates. We identify two substrates, namely metallic paint and blued metal, on which ridge lines become visible exclusively in UV range from 200 to 400 nm and from 210 to 300 nm correspondingly.

Hand-movement-based in-vehicle driver/front-seat passenger discrimination for centre console controls

Successful user discrimination in a vehicle environment may yield a reduction of the number of switches, thus significantly reducing costs while increasing user convenience. The personalization of individual controls permits conditional passenger enable/driver disable and vice versa options which may yield safety improvement. The authors propose a prototypic optical sensing system based on hand movement segmentation in near-infrared image sequences implemented in an Audi A6 Avant. Analyzing the number of movements in special regions, the system recognizes the direction of the forearm and hand motion and decides whether driver or front-seat passenger touch a control. The experimental evaluation is performed independently for uniformly and non-uniformly illuminated video data as well as for the complete video data set which includes both subsets. The general test results in error rates of up to 14.41% FPR / 16.82% FNR and 17.61% FPR / 14.77% FNR for driver and passenger respectively. Finally, the authors discuss the causes of the most frequently occurring errors as well as the prospects and limitations of optical sensing for user discrimination in passenger compartments.

Car-seat occupancy detection using a monocular 360° NIR camera and advanced template matching

The integration of seat occupancy detection systems is one of the most recent developments in automobile production. These systems prevent the deployment of airbags at unoccupied seats, thus avoiding the considerable cost imposed by the replacement of airbags. Seat-occupancy detection system can also be used to improve passenger comfort, e.g. by an occupation-dependent control of air-conditioning systems. This paper describes an inexpensive and versatile optical seat-occupancy detection system. Different approaches to pattern matching and the impact of local normalization, edge detection, multi-algorithm and temporal matching-score fusion are evaluated for each individual seat using a test set of 53,928 frames further classified in uniform and non-uniform illumination conditions. The results of these tests yield Equal Error Rates for uniform/non-uniform illumination of as low as 3.05%/1.68% for the front left seat, 2.17%/0.69% for the front right seat, 5.86%/4.01% for the rear left seat, 10.99%/11.07% for the rear center seat and 5.63%/1.84% for the rear right seat. The test results indicate that at least the two seat rows should be treated differently in terms of the selection of classification algorithms.