Tutorial 1
Title: Foundational GenAI Models in Biometrics – Full Day
Organizers: Mayank Vatsa (IIT Jodhpur) and Anush Sankaran (Microsoft Security AI Research)
Program: https://iab-rubric.org/tutorial-ijcb-2024
Abstract: The field of biometrics is crucial for various applications ranging from security to healthcare. Foundational AI models have significantly enhanced the accuracy, reliability, and application scope of biometric systems. This tutorial aims to cover the foundational aspects, recent advancements, and future trends in AI models applied to biometrics. Key topics will include introduction to foundational AI models (LLMs and LVMs), traditional biometrics techniques, transformative deep learning architectures, and cutting-edge research outcomes in AI. The session will also include several hands-on tutorials where we demonstrate how to solve some biometric problems using foundational AI models. The hands-on tutorial will be using Python notebooks.
Tutorial 2
Title: Biometric Privacy and Security – Half Day
Organizers: Vishnu Boddeti (MSU) and Amina Bassit (MSU)
Slides can be found from here
Abstract: Biometric systems have made tremendous progress over the past decade, thanks to the availability of copious data and computational resources and advances in deep learning methods. As we witness the widespread adoption of these systems, many privacy and security challenges now come to the fore. Biometric systems are capable of deeply analyzing and extracting hidden information that leads to learning specific human attributes. Consequently, the combination of those attributes can reveal one’s identity along with personal information merely from their processed biometric samples. The aim of this tutorial is to raise awareness among the biometric community about the alarming urgency of protecting biometric information from the risks its processing presents when it is conducted in an unprotected manner. This tutorial will discuss both privacy and security concerns through the results of existing studies and attacks. Then, it will discuss strategies to mitigate the concerns w.r.t. the different trade-offs to achieve an optimal equilibrium between efficiency, accuracy, security, and privacy. In addition, it will reflect on those solutions, identify gaps in current research, and motivate further research in those directions. Finally, through tangible examples, we provide a hands-on session that concretizes the biometric concepts and challenges discussed in the tutorial.
Tutorial 3
Title: Face Recognition Progression: Synthetic Images to Vulnerabilities – Half Day
Organizers: Akshay Agarwal (IISER Bhopal) and Chaitanya Roygaga (Lehigh University)
slides can be downloaded here
Abstract: Face recognition is one of the most commonly used technologies in the world, which involves finding and recognizing the identities of faces from images or videos by comparing them against an already available database of faces. The distribution of faces used for training recognition models might be different from each other, especially those with a degree of noise (corrupted) or those not captured in the physical environment (synthetic). The complexity of clean physical world images hinder effective feature learning. At a broad level, face swapping, morphing, and deepfake perform similar operations on the face images, but yield drastically different attack success rates — not only in fooling face recognition models, but also in attack detection algorithms. Further, the quality of face images and size of face images can significantly impact the features learned by the intermediate deep network layers. A model’s final performance, or the softmax output, could be a result of a combination of classifications (or misclassifications) across its various layers. This tutorial aims to shed light on these intriguing phenomena which might improve model explainability, by observing and describing the classification performance through the network’s layers. With an increase in data noise, deeper networks are generally preferred for a more robust model. We present a framework that allows feature visualization of input faces at multiple levels of a selected model, describing their correct (or incorrect) classification.
Tutorial 4
Title: Qualitative Methods for Biometrics Research: Exploring User Behavior and System Design – Half Day
Organisers: Tempestt Neal (University of South Florida)
Program: https://tempestt-neal.github.io/home/talks/ijcb24_tutorial
Abstract: Qualitative research is a method of inquiry aimed at gaining a deep understanding of social phenomena by relying on individuals’ direct experiences. Unlike quantitative research, which seeks to quantify variables and analyze numerical data, qualitative research emphasizes the exploration of complex, subjective experiences, meanings, and social dynamics. Qualitative exploration can greatly enhance the field of biometrics by offering deep insights into complex issues like bias in biometric systems and user acceptability. These methods allow for a more detailed understanding of how these systems are perceived and experienced, which is crucial for addressing ethical concerns and improving overall effectiveness. However, qualitative research design involves more than simply talking to people; it requires established strategies for systematically collecting, organizing, and interpreting data to understand people’s perspectives and experiences. This process demands proper researcher training and practice. This tutorial aims to provide biometrics researchers with a foundational understanding of qualitative research methods and their applicability to the field. This proposed half-day tutorial will cover foundational qualitative research concepts and their specific applications to biometrics. We will review qualitative research methodologies such as interviews, focus groups, case studies, and ethnographic studies. The tutorial will address qualitative research design including formulating research questions and hypotheses, selecting appropriate qualitative methods, sampling strategies for participant recruitment, and data collection techniques such as conducting and recording interviews, managing focus groups, and employing observational methods. Additionally, we will cover data analysis methods, including coding qualitative data, thematic analysis for identifying patterns and themes, and grounded theory for theory development. Ethical considerations in qualitative research, particularly regarding informed consent, confidentiality, and privacy in the context of biometric data, will also be discussed. We will explore how qualitative methods can enhance the understanding of user experiences related to biometric research, examine the social and cultural implications of biometric technologies, and investigate usability and acceptance across various contexts.