ERI: Towards Robust and Secure Intelligent 3D Sensing Systems

3D传感技术的进步是自动驾驶、身份验证和工业设计等现代应用不可或缺的一部分，它代表了我们与周围环境互动和理解能力的重大飞跃。通过使深度感知更容易获得和成本效益更高，这些技术为创新和实际应用开辟了新的途径，促进了国家的繁荣和福祉。然而，随着这些技术越来越多地融入我们的日常生活，它们所支持的系统的安全性已经成为一个关键问题。开发强大而安全的智能3D传感系统对于网络物理系统和基础设施的可靠运行至关重要。该项目旨在通过开发智能3D传感的安全框架来解决这些挑战，重点关注安全性和可靠性至关重要的自动驾驶汽车等应用。其目的是保护这些技术免受潜在威胁，确保它们对社会进步和科学的安全进步作出积极贡献。该项目旨在通过发表学术文章和向更广泛的社区发布开源软件、示范资源和数据集来分享其见解和突破。将为本科生和研究生提供研究培训机会，并积极努力吸引代表性不足的群体的候选人。通过这些努力，该项目将提高我们对3D传感技术的理解和能力，并为新一代学生和工程师提供知识和工具，以应对这一快速发展领域的未来挑战。该项目解决了围绕三个相互关联的研究任务构建的智能3D传感系统相关的新兴安全问题。第一个任务是深入了解3D深度学习领域中分布外（OOD）样本的独特性质。与2D样品不同，3D样品具有不同的特征和分布，需要进行彻底的调查，以减轻其对系统可靠性的潜在影响。第二个任务侧重于开发一个全面的框架，以增强感知管道的各个方面和阶段的鲁棒性。该框架旨在防止物理对抗性威胁，确保关键环境中3D传感应用的安全性。最后的任务是评估阶段，包括创建仿真工具和测试平台。这些工具将通过对现有和新收集的数据集进行广泛测试，并在实际场景中实施，验证前两项任务中开发的发现和方法。这种全面的方法确保了该项目不仅解决了理论挑战，而且在实际应用中验证了其发现，为在许多环境中部署安全、智能的3D传感系统铺平了道路。通过将严谨的研究与实际验证相结合，该项目旨在为智能3D传感做出重大贡献，提高其可靠性和安全性。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The advancement of 3D sensing technology, integral to modern applications like autonomous driving, identity verification, and industrial design, represents a significant leap forward in our ability to interact with and understand the environment around us. By making depth perception more accessible and cost-effective, these technologies open new avenues for innovation and practical application, enhancing national prosperity and welfare. However, as these technologies become more embedded in our daily lives, the security of the systems they enable has emerged as a critical concern. The quest for developing robust and secure intelligent 3D sensing systems becomes crucial for the reliable functioning of cyber-physical systems and infrastructure. This project seeks to address these challenges by developing a secure framework for intelligent 3D sensing, focusing on applications such as autonomous vehicles where safety and reliability are paramount. The aim is to safeguard these technologies from potential threats, ensuring they contribute positively to societal advancement and the secure progression of science. This project aims to share its insights and breakthroughs by publishing scholarly articles and releasing open-source software, demonstration resources, and datasets to the broader community. Research training opportunities will be provided to both undergraduate and graduate students, with proactive efforts to attract candidates from underrepresented groups. Through these efforts, the project will advance our understanding and capabilities in 3D sensing technologies and prepare a new generation of students and engineers equipped with the knowledge and tools to tackle future challenges in this rapidly evolving area.This project addresses the emerging security concerns associated with intelligent 3D sensing systems structured around three interlinked research tasks. The first task delves into understanding the unique nature of out-of-distribution (OOD) samples in the 3D deep learning domain. Unlike their 2D counterparts, OOD samples in 3D possess distinct characteristics and distributions that require thorough investigation to mitigate their potential impact on system reliability. The second task focuses on developing a comprehensive framework that enhances robustness across various aspects and stages of the perception pipeline. This framework aims to protect against physical adversarial threats, ensuring the security of 3D sensing applications in critical environments. The final task, the evaluation phase, involves the creation of simulation tools and testbed platforms. These tools will validate the findings and methodologies developed in the first two tasks through extensive testing with existing and newly collected datasets and implementation in real-world scenarios. This comprehensive approach ensures that the project not only addresses theoretical challenges but also validates its findings in practical applications, paving the way for deploying secure, intelligent 3D sensing systems in many settings. By integrating rigorous research with practical validation, this project aims to significantly contribute to intelligent 3D sensing, enhancing its reliability and security.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.