Collaborative Research: SaTC: EDU: Adversarial Malware Analysis - An Artificial Intelligence Driven Hands-On Curriculum for Next Generation Cyber Security Workforce

人工智能（AI）和机器学习（ML）技术可以通过帮助安全管理员检测可疑行为和启动对威胁的响应来加强网络安全。然而，AL/ML技术仍然容易受到恶意利用，可能导致意想不到的结果。因此，重要的是要确保基于人工智能的决策过程在面临对抗性情况时在关键操作系统中是可靠的。随着深度学习（DL）和其他AI/ML算法被集成到操作系统中，必须保护AI/ML的安全性、隐私性和公平性。这可以通过实施更强大的ML方法来实现，例如AI侦察预防，对抗模型分析，模型中毒预防和安全训练程序。通过为学生提供在恶意软件分析应用程序中保护人工智能所需的知识，该项目将促进下一代网络安全人才的增长。该项目将在恶意软件分析应用程序的背景下研究和开发专注于对抗性机器学习（AML）的独立课程模块，这将把前沿研究课题转化为教学和学习过程。这些模块的目标是培养学生在田纳西理工大学（TTU）和北卡罗来纳州农业和技术州立大学（NCAT）在这一领域的专业知识。课程模块将包括对抗性恶意软件生成，文件结构对随机扰动的鲁棒性，中毒攻击和防御，白盒规避攻击和代理模型构建。AML网络模块将被整合到不同的非安全课程中，如AI/ML或数据科学，或作为独立的网络安全课程提供。学生将通过参与不同的AI/ML技术来获得与恶意软件分析领域相关的安全解决方案的实践和概念知识。此外，学生将开发保护AI系统所需的高级技能。由网络安全、人工智能和教育专家组成的跨学科团队将利用指导性概念框架，从战略上开发网络安全教育模块。他们将调查这些模块对学习成果的影响，同时完善教学策略，以促进网络安全教育的多样性和包容性。将广泛提供已开发的单元、教学材料和辅导活动，以供传播。该项目将支持安全和教育研究课题的整合，以创造网络安全方面的新知识。该奖项反映了NSF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。

Artificial Intelligence (AI) and Machine Learning (ML) techniques can bolster cybersecurity by aiding security administrators in detecting suspicious behaviors and initiating responses to threats. However, AL/ML technology remains susceptible to malicious exploitation, potentially leading to unintended outcomes. Therefore, it is important to ensure that AI-based decision processes are reliable in critical operational systems when facing adversarial situations. As deep learning (DL) and other AI/ML algorithms become integrated into operational systems, it is essential to defend security, privacy, and fairness of AI/ML against adversaries. This can be achieved by implementing more robust ML methods such as AI reconnaissance prevention, analysis of adversarial models, model poisoning prevention, and secure training procedures. By equipping students with the knowledge needed to secure AI in malware analysis applications, this project will foster growth of next-generation cybersecurity talent. This project will research and develop self-contained course modules focused on Adversarial Machine Learning (AML) within the context of malware analysis applications, which will transit cutting-edge research topics into the teaching and learning process. The goal of these modules is to develop students at Tennessee Tech University (TTU) and North Carolina Agricultural and Technical State University (NCAT) with specialized knowledge in this area. Course modules will include adversarial malware generation, robustness of file structure against random perturbation, poisoning attack and defense, white-box evasion attack, and surrogate model construction. The AML cyber modules will be integrated into different non-security courses such as AI/ML or data science or provided as an independent cybersecurity course. Students will acquire practical and conceptual knowledge by engaging with different AI/ML techniques for security solutions pertinent to the malware analysis domain. Additionally, students will develop advanced skills necessary for safeguarding AI systems. The interdisciplinary team, composed of experts in cybersecurity, artificial intelligence, and education, will utilize a guiding conceptual framework to strategically develop cybersecurity education modules. They will investigate the impact of these modules on learning outcomes, while refining pedagogical strategies to promote diversity and inclusion in cybersecurity education. Developed modules, instructional materials, and tutorial activities will be widely available for dissemination. This project will support integration of security and education research topics to create new knowledge in cybersecurity.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.