ABSTRACT. The use of biometric authentication on mobile devices such and smartphones and tablets represents the largest public use of the security technology. Whilst levels of operational performance have enabled trust in the technology, research questions still remain as to implementation and optimisation using facets such as novel sensors. This presentation provides and overview of the research and application of biometrics on mobile devices and will look to future applications and research questions yet to be answered.

ABSTRACT. Safety in industrial robotic environments is a hot research topic in the area of human-robot interaction (HRI). Up to now, a robotic arm on an assembly line interacts with other machines away from human workers. Nowadays, robotic arm manufactures are aimed to their robots could increasingly perform tasks collaborating with humans. One of the ways to improve this collaboration is by making the movement of robots more humanlike. This way, it would be easier for a human to foresee the movement of the robot and approach it without fear of contact. The main difference between the movement of a human and of a robotic arm is that the former has a bell-shaped speed profile while the latter has a uniform speed one. To generate this speed profile, the kinematic theory of rapid human movements and its Sigma-Lognormal model has been used. This model is widely used to explain most of the basic phenomena related to the control of human movements. Both human-like and robotic-like movements are transferred to the UR3 robot. In this paper we detail the how the UR3 robot was programmed to produce both kinds of movement. The dissimilarities result between the input motion and output motion to the robot confirm the possibility to develop human-like velocities in the UR3 robot.

ABSTRACT. One of the direct solutions to tackle the spread of the COVID-19 virus is the usage of face masks, with face mask wearing being compulsory in public places for everyone. Monitoring of the compliance of citizens with this measure is essential for the immediate detection of cases of people non wearing face masks, which will allow the control of potential spread of the virus. Face mask detection can be automated by using cameras and computer vision methods. In this paper we evaluate the performance of different deep learning models in terms of face detection accuracy as well as implementations of them on hardware boards measuring the corresponding inference time.

ABSTRACT. Verifying the authorship of a questioned signature is a common task for forensic handwriting examiners. While the automatic systems are typically orientated to improve performance, their practical utility for forensics is not always guaranteed. In this paper, we propose an offline automatic signature verifier oriented to forensic handwriting examiners. Our design is based on likelihood ratios which translate the signature verification results into objective and understandable evidence for a jury in a courtroom. The likelihood ratios depend on a universal background model build with signatures from other users and distance measures between signature handcrafted features. These features can be more interpretable for forensics, even though others can be included in our verifier, like deep learning ones. In our experiments, a single signature was used as reference. Two universal background models have been developed -- the first is based on the GPDS database, and the second on synthetic signatures. The scheme is tried and tested with signatures from MCYT75 and BiosecureID databases with promising results. The outcome of this work is an offline signature verifier for forensic handwriting examiner practice.

ABSTRACT. Significant advances in deep learning have obtained hallmark accuracy rates for various computer vision applications. However, advances in deep generative models have also led to the generation of very realistic fake content, also known as deepfakes, causing a threat to privacy, democracy, and national security. Most of the current deepfake detection methods are deemed as a binary classification problem in distinguishing authentic images or videos from fake ones using two-class convolutional neural networks (CNNs). These methods are based on detecting visual artifacts, temporal or color inconsistencies produced by deep generative models. However, these methods require a large amount of real and fake data for model training and their performance drops significantly in cross dataset evaluation with samples generated using advanced deepfake generation techniques. In this paper, we thoroughly evaluate the efficacy of deep face recognition in identifying deepfakes, using different loss functions and deepfake generation techniques. Experimental investigations on challenging Celeb-DF and FaceForensics++ deepfake datasets suggest the efficacy of deep face recognition in identifying deepfakes over two-class CNNs and the ocular modality. Reported results suggest a maximum Area Under Curve (AUC) of 0.98 and Equal Error Rate (EER) of 7.1% in detecting deepfakes using face recognition on the Celeb-DF dataset. This EER is lower by 16.6% compared to the EER obtained for the two-class CNN and the ocular modality on the Celeb-DF dataset. Further on the FaceForensics++ dataset, an AUC of 0.99 and EER of 2.04% was obtained. The use of biometric facial recognition technology has the advantage of bypassing the need for a large amount of fake data for model training and obtaining better generalizability to evolving deepfake creation techniques.

ABSTRACT. A number of studies suggest bias of the face biometrics, i.e., face recognition and soft-biometric estimation methods, across gender, race, and age-groups. There is a recent urge to investigate the bias of different biometric modalities toward the deployment of fair and trustworthy biometric solutions. Ocular biometrics has obtained increased attention from academia and industry due to its high accuracy, security, privacy, and ease of use in mobile devices. A recent study in $2020$ also suggested the fairness of ocular-based user recognition across males and females. This paper aims to evaluate the fairness of ocular biometrics in the visible spectrum among age-groups; young, middle, and older adults. Thanks to the availability of the latest large-scale 2020 UFPR ocular biometric dataset, with subjects acquired in the age range 18 - 79 years, to facilitate this study. Experimental results suggest the overall equivalent performance of ocular biometrics across gender and age-groups in user verification and gender-classification. Performance difference for older adults at lower false match rate and young adults was noted at user verification and age-classification, respectively. This could be attributed to inherent characteristics of the biometric data from these age-groups impacting specific applications, which suggest a need for advancement in sensor technology and software solutions.

ABSTRACT. Biometric authentication is nowadays widely used in a multitude of scenarios. Several studies have been conducted on electrocardiogram (ECG) for subject identification or verification among the various modalities. However, none have considered a typical implementation with a mobile device and the necessity for a fast-training model with limited recording time for the signal. This study tackles this issue by exploring various classification models on short recordings and evaluating the performance varying the sample length and the training set size. We run our tests on two public datasets collected from wearable and medical devices and propose a pipeline for ECG authentication with limited data required for competitive usage across applications.

ABSTRACT. In the past decades, there has been a tremendous growth in the security-related issues throughout the world. With such a multifold increase in demand for security, video surveillance has become one of the important research areas. In the most recent studies, there were many attempts to integrate image processing algorithms and was possible to achieve artificial intelligence capabilities using computer vision. The capabilities like prevention, detection and intervention have led to the development of real and consistent video surveillance systems with intelligent processing techniques. In other words, intelligent video surveillance systems can be described as a technique to assist security by providing reliable real-time alerts and to support the investigations with efficient video analysis. In this field of research, there has been a large amount of datasets, methods and algorithms published, but there are not many publications which can provide an entire picture of the current state-of-the-art video surveillance systems. This paper provides a comprehensive and systematic overview on the existing surveillance techniques and possible enhancements.

ABSTRACT. Advances in machine learning mean that automated decision systems are increasingly interpreting images that were previously only seen by people. In critical applications like security imaging, it is imperative that automatic target recognition (ATR) algorithms behave predictably and robustly. However, such algorithms often perform poorly when image quality changes due to altered environmental conditions, excessive vibration, or gradual system degradation, so it is important to detect such changes. For this purpose, we have developed tools for real-time image quality estimation based on natural scene statistics (NSS). The real-time methods leverage statistical measurements intrinsic to natural images to assess the similarity between an input image and a set of training images. We demonstrate the tool for a millimeter-wave (MMW) imager used for detection of concealed contraband on screened people. We also discuss the use of similar methods for real-time monitoring of x-ray imagery.

TALK ONLY, NO PAPER