Biometric recognition of children presents unique challenges due to rapid physiological changes that affect the consistency of extracted features over time. This study presents a longitudinal evaluation of a multimodal biometric system combining facial and ear features for child recognition across a three-year span. The evaluation uses collected datasets containing frontal and profile images of 231 children aged 3 to 18, with seven data collection sessions spaced at six-month intervals. Face recognition was performed using MagFace, while ear recognition was based on a pipeline involving Mask R-CNN for segmentation and an ensemble of VGG16 and MobileNet for feature extraction. Independent evaluations showed an increase in Equal Error Rate (EER) for face from 1.20% to 2.55% over a 36 -month interval, and for ear from $5.62 \%$ to $19.00 \%$ over the same period. When the two modalities were fused at the feature level, the system achieved improved stability with EER values ranging from $0.24 \%$ at 6 months to $1.38 \%$ at 36 months. Additional analysis across age groups revealed that children enrolled at younger ages experienced higher error rates, highlighting the importance of age-aware system design. These results demonstrate the effectiveness of face-ear fusion in supporting inclusive and temporally stable biometric systems, particularly for applications in education, healthcare, and identity tracking in humanitarian contexts.

Face recognition for infants and toddlers presents unique challenges due to rapid facial morphology changes, high inter-class similarity, and the limited availability of datasets. This study evaluates the performance of four deep learning-based face recognition models—FaceNet, ArcFace, Mag-Face, and CosFace—on a newly developed longitudinal dataset collected over a 24-month period in seven sessions involving children aged 0 to 3 years. Our analysis investigates recognition accuracy across multiple developmental stages, showing that the True Accept Rate (TAR) is only 30.7% at 0.1% False Accept Rate (FAR) for infants aged 0 to 6 months due to unstable facial features, but improves significantly in older children, reaching 64.7% TAR at 0.1% FAR in the 2.5 to 3 year age group. We also examine how face verification performance changes in different time intervals, revealing that shorter time gaps produce better accuracy due to reduced embedding drift. To mitigate this drift, we apply Domain-Adversarial Neural Network (DANN) strategy that improves TAR by more than 12% and yields features that are more temporally stable and generalizable. These findings are critical for building biometric systems that function reliably over time in smart city applications such as public healthcare, child safety, and digital identity services. The challenges observed in early age groups also highlight the importance of future research on privacy-preserving biometric authentication systems that can address temporal variability, especially in secure and regulated urban environments where child verification is vital.

LivDet-Iris 2025 is the sixth edition of the iris liveness detection competition. Held every two to three years, the competition aims to foster the development of robust algorithms capable of detecting a wide range of physically-and digitally-presented attacks in iris biometrics. The 2025 edition obtained the largest number of submissions in the history of the competition: ten algorithms from five institutions, and one commercial iris recognition system. LivDet-Iris 2025 also introduced new tasks compared to previous editions: (Task 1) a benchmark offered by an industry partner, (Task 2) morphed iris images, in which two different-identity samples were blended into one image, and (Task 3) evaluation of presentation attack detection robustness against advanced manufacturing techniques for textured contact lenses. This edition, for the first time in the series, offers a systematic testing of a commercial iris recognition system (software and hardware) using physical artifacts presented to the sensor. Dermalog-Iris team submitted algorithms that won all tasks, achieving the area under the ROC curve of 90.57%, 68.23% and 99.99% in tasks 1, 2, and 3, respectively. Additionally, we include results for baseline algorithms, based on modern deep convolutional neural networks and trained with all available public datasets of iris images representing bona fide samples and anomalies (physical attacks, eye diseases, post-mortem cases, and synthetically-generated iris images). Test samples created for tasks 2 and 3, and baseline models are made available to offer the state-of-the-art benchmark for iris liveness detection.

Imagine a world where a copy of your face could trick the most advanced security systems. This isn’t science fiction; it’s a real challenge today. LivDet-Face is a competition that aims to advance the detection of attacks at the biometric sensor, known as Presentation Attack Detection (PAD). This international contest is a key benchmark in biometric security, offering an unbiased look at the latest innovations in face PAD and demonstrating progress over time in detecting and preventing sophisticated attacks. Through the International Joint Conference on Biometrics (IJCB) platform, LivDet-Face 2024 provides a standardized evaluation process, access to advanced Presentation Attack Instruments (PAI), and a comprehensive dataset of bona fide face images. The competition had two main categories: algorithms and systems. A total of sixteen algorithms and one system were submitted for this year’s competition. Anonymous submissions topped both image and video subcategories with an ACER of 4.93% and 4.13%, respectively. In the systems category, Team Dermalog, despite being the sole submission, achieved an impressive ACER of 3.12%.

A longitudinal cohort study is a popular research method to observe a group of people over a prolonged period of time, e.g., to learn about their health, wellness, and social habits. Smartphones have become a very popular tool to perform such studies at a large scale. Location is an essential form of sensor data that can not only be used to monitor users' mobility and social interaction patterns, but also to identify places of personal significance, i.e., places where a user spends a significant amount of time, such as a user's home, workplace, and preferred social gathering places. However, continuously tracking a user's location can have significant impacts on the battery lifetime of a smartphone. Therefore, instead of frequent period location sensing, this paper identifies smartphone events that can be used to trigger location sensing at a much lower rate (and therefore more energy-efficiently), while still providing accurate location data. In this work, we demonstrate that this approach allows us to determine a user's significant places with an accuracy of 85%, while saving over 60% in computational and energy overheads.

Continuous (24/7) smartphone sensing applications are on the rise, especially in the field of health and wellness, e.g., to monitor physical activity, to quickly detect emergencies (e.g., falls), and to provide various context-specific services and tools. Smartphones are able to monitor a large array of human activities and patterns, including a user's mobility, physical activities, social interactions, mobile app usage, or communication events. However, building smartphone sensing applications that operate reliably and efficiently on a continuous basis is challenging. Specifically, in this paper, we describe our experience with building a sensing service for the iOS platform, that has been running on more than 400 devices continuously for more than 9 months to date. We present and discuss a variety of technical and non-technical obstacles and challenges and how they were addressed.

Wide-issue superscalar processors have capabilities to execute several basic blocks in a cycle. A regular instruction cache fetch mechanism is not capable of supporting this high fetch throughput requirement. Several improvements of the fetch mechanism are currently in use. One of the most successful of these improvements is the addition of an instruction memory structure known as a trace cache. In this paper an analytical model of instruction fetch performance of a trace cache microarchitecture is presented. Parameters, which affect trace cache instruction fetch performance, are explored and several analytical expressions are presented. The presented model can be used to understand performance tradeoffs in trace cache design. Results from the validation of the model are presented. The instruction fetch rates predicted by the model differ by seven percent from the simulated fetch rates for SPEC2000 benchmark programs. The model is implemented in a computer program named Tulip. To show how different parameters influence performance, results from Tulip are also presented.

Instruction fetch mechanism is a performance bottleneck of a Superscalar Processor. The fetch performance of the processor can be improved with the aid of an instruction memory structure known as Trace Cache. This paper presents parameters and analytical expressions, which describe instruction fetch performance of a Trace Cache microarchitecture. The instruction fetch rates predicted by the expressions differ by seven percent from the simulated fetch rates for SPEC2000 benchmark programs. Presented analytical expressions are implemented in a computer program named Tulip. Tulip is used to explore parameters, and their influence on fetch performance. Tulip is also used to understand Trace Cache performance tradeoffs.

This paper presents an analytical model of instruction fetch performance of a trace cache. This paper also presents an analytical model of miss rate of a trace cache. These models can be used to analyze performance and behavior of a microarchitecture of a processor. These models are implemented in a new microarchitecture tool Tulip. Performances of several benchmark programs based on Tulip are also presented in this paper.