Special Session #1: Generative AI for Futuristic Biometrics (Genai-2024)
Generative AI has significantly reshaped modern machine learning in both vision and language domains in terms of unprecedented realism, diversity and efficiency. This will also have an impact on biometric research that heavily relies on large-scale, diverse, sensitive and personally identifiable data. On one hand, we can leverage generative models for controllable synthesis of large amounts of biometric data in an efficient automated fashion. This synthetic data in turn can be used for de-biasing existing models or extending data when real data is limited. This is essential, specifically, in extreme cases such as post-mortem-based recognition or recognition of infants and children, where data collection is significantly restricted or nearly impossible. Another potential use of generative models can be adaptation of text-driven large language models to produce natural language interpretation of data. The generated descriptions can explain the decisions generated by the biometric systems thereby, making them more trustworthy and explainable. On the other hand, generative AI can be used in an adversarial capacity to circumvent existing systems. Novel attack vectors such as spoofs, template inversion and Deepfake can be simulated more effectively using generative AI. Biometrics of the future should therefore, utilize this novel potential of generative AI to both identify vulnerabilities in existing systems and develop intelligent, trustworthy and robust systems.
Website: https://sites.google.com/nyu.edu/genai-fb/home/committee?authuser=0
Special Session #2: Recent Advances in Detecting Manipulation Attacks on Biometric Systems (ADMA-2024)
Manipulated attacks in biometrics via modified images/videos and other material-based techniques such as presentation attacks and deep fakes have become a tremendous threat to the security world owing to increasingly realistic spoofing methods. Hence, such manipulations have triggered the need for research attention towards robust and reliable methods for detecting biometric manipulation attacks. The recent inclusion of manipulation/generation methods such as auto-encoder and generative adversarial network approaches combined with accurate localisation and perceptual learning objectives added an extra challenge to such manipulation detection tasks. Due to this, the performance of existing state-of-the-art manipulation detection methods significantly degrades in the unknown scenarios. Apart from this, real-time processing, manipulation on low-quality medium, limited availability of data, and inclusion of these manipulation detection techniques for forensic investigation are yet to be widely explored. Hence, this special session aims to profile recent developments and push the border of the digital manipulation detection technique on biometric systems.
Website: https://sites.google.com/view/ss-adma2024/home
Special Session #3: Face Morphing Attack and Detection Techniques (FMADT-2024)
Face morphing attacks have emerged as a potent attack vector targeting state-of-the-art Face Recognition (FR) systems. FR, which should be tolerant with respect to intra-class variations by design, turn out to be vulnerable to such attacks. Designing algorithms to detect this emerging threat is of preeminent relevance to secure FR systems deployed across a wide range of operational applications. However, the success of developing effective Morphing Attack Detection (MAD) algorithms in a rapidly evolving landscape against synthetic (and non-synthetic in some cases) image generation technology will be highly dependent on access to the latest morph generation technology, methods, and data. By developing more openly accessible morph generation algorithms and datasets, we enable the research community to train their MAD algorithms on the most potent and effective morphing algorithms, shutting down potential attack vectors. Lastly, recent work has shown that the post-processing and the medium, i.e., printed and scanned images or purely digital images, of both the suspected image and the trusted live captured image can greatly impact the efficacy of the morphed attack. Towards this aim we invite researchers to submit papers towards this special session at IJCB 2024 under the general envelope of face morphing attack and detection techniques.
Website: https://sites.google.com/clarkson.edu/fmadt2024/home
Special Session #4: Recognition at Long Range and from High Altitude (LRR-2024)
Biometric recognition from imagery has been studied for several decades, and the frontier of recognition capabilities has expanded with the development of underlying computational tools. Deep neural networks, for instance, have enabled robust face recognition from close ranges and from viewpoints commonly represented in web-scale face image datasets. With increased data resources and funding sources, an area of emphasis has developed around recognition from imagery captured at long ranges and high altitudes. These situations are characterized by challenges such as atmospheric turbulence, occlusion, and non-traditional viewpoints.
Website: https://sites.google.com/kitware.com/ijcb-lrr-session-2024/home
Special Session #5: Multimodal Human Behavior Understanding and Generation (MUG-2024)
Human behavior involves not only language expression, but also facial expressions, body movements, voice tone, and other modalities. Understanding and simulating human behavior requires the integration of these multimodal information, rather than relying on a single modality. Going deeper in field of Multimodal Human Behavior Understanding and Generation (MUG) can benefit the deep understanding multimodal nature of human behavior. Furthermore, Multimodal understanding and generation of human behavior can help computer systems better perceive, understand and respond to human intentions and emotional states. This can make human-machine interaction more natural and smooth, thereby enhancing user experience. Hence, this special session aims to profile recent developments on multimodal biometric system, especially on trustworthy multimodal data integration, ognitive and neurological underpinnings, generative modeling on human behavior, and potential in board real-world applications.
Website: https://sites.google.com/view/ijcb-mug2024
Special Session #6: Responsible AI for Biometrics (AI4BIO)
Responsible AI for Biometrics (AI4BIO) is critical and timely, given the rapid expansion and integration of biometric technologies into various sectors such as security, finance, healthcare, and consumer electronics. The ethical deployment of these technologies is crucial to avoid potential misuse, discrimination, and violation of privacy and human rights. The special session covers topics ranging from accountability and fairness to privacy and security, which address comprehensive and crucial issues that ensure biometric technologies are developed and used in a manner that benefits society without compromising ethical standards. The significance of these topics lies in their holistic approach to the responsible integration of biometrics in society. Each topic, such as equity, inclusion, and sensitivity to culture and context, acknowledges the profound implications biometric technologies have on diverse populations. The novelty comes from addressing these issues in a combined manner, offering a multifaceted view that is often missing in more technically focused discussions. This comprehensive approach ensures that technological advancements enhance societal well-being and do not perpetuate or exacerbate existing disparities.
Website: https://sites.google.com/view/ai4bio2024
Special Session #7: Face Recognition in Era of Synthetic Images and Its Boundless Vulnerabilities (SIBV-SS)
The vulnerabilities of face recognition algorithms are limitless; hence this special session covers a wide range of topics that highlight the positive and negative aspects of the factors affecting face recognition. The topics include deepfake, the use of synthetic media for privacy-preserving learning, facial attribute annonymization, adversarial attacks, morphing, and presentation attacks. Face recognition has been proven one of the most effective for establishing identities; however, the malicious purposes of intruders and the advancement of automated technologies have led to the development of several anomalies that can trick the system. However, the literature rarely describes these different anomalies under one roof, which limits the understanding of the functioning of the different anomalies or features that might not be an adversary but are used as an adversary due to poor network learning. This session aims to provide a comprehensive understanding of the success of face recognition algorithms and how different factors contribute to their success such as synthetic images or failure such as adversarial attacks. We assert due to the involvement of the significant inter-disciplinary concept, the proposal can help in understanding face recognition from a top level. For example, the generation of deepfake and adversarial attacks is significantly different but in the end, they are manipulating the deep-level features of deep face recognition. Understanding how these factors are working can help us in developing a universally robust deep face recognition. The proposed special session is critical and highly relevant to the audience of the main conference; therefore, we request the community to actively take part and submit their high-quality papers to understand and protect the integrity of deep face recognition networks.
Website: https://sites.google.com/iiserb.ac.in/ijcb24-sibv-ss/home?authuser=0