EAGER: SaTC-EDU: Privacy Enhancing Techniques and Innovations for AI-Cybersecurity Cross Training

EAGER：SaTC-EDU：人工智能-网络安全交叉培训的隐私增强技术和创新

• 批准号: 2038029
• 负责人: Ling Liu
• 金额: $ 30万
• 依托单位: Georgia Tech Research Corporation
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2020
• 资助国家: 美国
• 起止时间: 2020-09-01 至 2023-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2038029&HistoricalAwards=false
• 关键词: EAGER; SaTC; EDU; Privacy; Enhancing

Artificial intelligence (AI) is being rapidly deployed in many security-critical applications. This has fueled the use of AI to improve cybersecurity via speed of reasoning and reaction (AI for cybersecurity). At the same time, the widespread use of AI introduces new adversarial threats to AI systems and highlights a need for robustness and resilience guarantees for AI (cybersecurity for AI), while ensuring fairness of and trust in AI algorithmic decision making. Not surprisingly, privacy-enhancing technologies and innovations are critical to mitigating the adverse effects of intentional exploitation and protecting AI systems. However, resources for AI-cybersecurity cross-training are limited, and even fewer programs integrate topics, techniques and research innovations pertaining to privacy in their basic curricula covering AI or cybersecurity. To bridge this cross-training gap and to advance AI-cybersecurity education, this project will create a pilot program on privacy-enhancing AI-cybersecurity cross-training, which will provide a transformative learning experience for students. The results of this project will provide students with the AI-cybersecurity knowledge and skills that will enable them to enter the workforce and contribute to the creation of a secure and trustworthy AI-cybersecurity environment that simultaneously supports AI safety, AI privacy and AI fairness for all. The intellectual merit of this project stems from the development of a first-of-its-kind research and teaching methodology that will provide effective AI-cybersecurity cross-training in the context of privacy. This will include developing a privacy foundation virtual laboratory (vLab) and three advanced topic vLabs, each representing a unique educational innovation for AI-cybersecurity cross-training. The AI for Security vLab will enable students to learn that privacy is a critical system property for all AI-enabled cybersecurity systems and applications. The Security of AI vLab will assist students in learning that privacy is an important safety guarantee against a variety of privacy leakage risks. The AI Fairness and Trust vLab will empower students to learn that privacy is an essential measure of trust and fairness of AI systems by ensuring the right to privacy and AI ethics for all. By participating in these vLabs, students will learn to use risk assessment tools to understand new vulnerabilities to attack of AI models and to design risk-mitigation tools to protect AI model learning and reasoning against security or privacy violations and algorithmic biases.This project is supported by a special initiative of the Secure and Trustworthy Cyberspace (SaTC) program to foster new, previously unexplored, collaborations between the fields of cybersecurity, artificial intelligence, and education. The SaTC program aligns with the Federal Cybersecurity Research and Development Strategic Plan and the National Privacy Research Strategy to protect and preserve the growing social and economic benefits of cyber systems while ensuring security and privacy.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.

人工智能（AI）正在迅速部署在许多关键安全应用程序中。这加剧了通过推理和反应速度（网络安全的AI）来改善网络安全的使用。同时，广泛使用AI对AI系统引入了新的对抗性威胁，并突出了对AI（AI的网络安全）的鲁棒性和弹性保证的需求，同时确保了AI算法的公平性和信任。毫不奇怪，增强隐私技术和创新对于减轻故意开发和保护AI系统的不利影响至关重要。但是，AI-Cyber​​security交叉培训的资源有限，甚至更少的计划将与隐私有关的主题，技术和研究创新整合到涵盖AI或网络安全的基本课程中。为了弥合这个交叉训练的差距并提高AI-Cyber​​security教育，该项目将创建一个有关增强隐私的AI-Cyber​​bersecurity交叉培训的试点计划，该计划将为学生提供变革性的学习经验。该项目的结果将为学生提供AI-Cyber​​security的知识和技能，使他们能够进入劳动力，并为创建一个安全且值得信赖的AI-Cyber​​berity环境做出贡献，同时支持AI安全，AI的隐私和AI公平性。该项目的智力优点源于开发首个研究和教学方法，该方法将在隐私的背景下提供有效的AI-Cyber​​security交叉培训。这将包括开发一个隐私基金会虚拟实验室（VLAB）和三个高级主题VLABS，每个主题vlabs代表了AI-Cyber​​bersecurity交叉培训的独特教育创新。安全VLAB的AI将使学生能够了解隐私是所有支持AI支持网络安全系统和应用程序的关键系统属性。 AI VLAB的安全性将帮助学生了解隐私是针对各种隐私泄漏风险的重要安全保证。人工智能公平和信任VLAB将通过确保所有人的隐私权和AI道德权，使学生能够学习隐私是对AI系统信任和公平性的基本衡量。通过参与这些VLAB，学生将学习使用风险评估工具来了解攻击AI模型的新漏洞，并设计风险降低的工具，以保护AI模型学习和推理违反安全或隐私的偏见以及算法的偏见。该项目由该项目提供了一项特殊和信用的网络（以前是New of New of Satcace），以前的领域（以前），以前是New satc satce of satc of satc of satc of satc ostpl ostpl ostpl ostplost forppl ostpl oster forploster forploster for ost forppl ostplospl网络安全，人工智能和教育。 SATC计划与联邦网络安全研究与发展战略计划以及国家隐私研究策略保持一致，以保护和保留网络系统的社会和经济益处，同时确保安全和隐私。该奖项反映了NSF的法定任务，并认为通过基金会的知识分子和更广泛的影响，可以通过评估来进行评估，以审查Criteria。

项目成果

Metric Learning as a Service With Covariance Embedding

• DOI: 10.1109/tsc.2023.3266445
• 发表时间: 2022-11
• 期刊: IEEE Transactions on Services Computing
• 影响因子: 8.1
• 作者: Imam Mustafa Kamal;Hyerim Bae;Ling Liu

Boosting Object Detection Ensembles with Error Diversity

• DOI: 10.1109/icdm54844.2022.00105
• 发表时间: 2022-11
• 期刊: 2022 IEEE International Conference on Data Mining (ICDM)
• 作者: Ka-Ho Chow;Ling Liu

Selecting and Composing Learning Rate Policies for Deep Neural Networks

选择和制定深度神经网络的学习率策略

• DOI: 10.1145/3570508
• 发表时间: 2023
• 期刊: ACM Transactions on Intelligent Systems and Technology
• 影响因子: 5
• 作者: Wu, Yanzhao;Liu, Ling
• 通讯作者: Liu, Ling