Assistant Professor

助理教授

• 批准号: RGPIN-2020-06962
• 负责人: Ding, StevenHonghui
• 金额: $ 2.11万
• 依托单位: Queen's University
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2020
• 资助国家: 加拿大
• 起止时间: 2020-01-01 至 2021-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=718577
• 关键词: Assistant; Professor

With the introduction of ultra-fast 5G networks, companies and organizational infrastructures are facing an ever-increasing number of endpoint devices such as PCs, mobile devices, Internet-of-Thing (IoT) sensors, and actuators. They significantly enlarge the attack surface, and the cyber-attacks have been vastly shifting from perimeter to endpoint devices. Malware infection and fileless vulnerability exploits are two major rapidly evolving threats against endpoints, causing significant financial loss to companies and organizations. Deep Learning (DL) has been driving the next-generation malware detection and vulnerability discovery solutions for optimized endpoint security in both industry and academia, due to its simplicity for integration, low overhead, and effectiveness against future unknown attacks. However, these DL-powered solutions as black-box approaches cannot provide the same degree of insight and actionable information as legacy signature-based solutions to security analysts and officers. The proposed research addresses the critical and urgent issues of effectiveness, interpretability, and actionability of the black-box DL solutions against emerging malware and vulnerability exploits on endpoint devices. The research studies the underlying generic representation of malware and vulnerability in a software ecosystem and their relationship to build an effective DL system defending against and gaining insight from the incoming attacks. This research will be fundamental to promote Canadian cyber capability to detect, defend, and act on emerging large-scale cyber-attacks targeting both the public and private sectors. It also augments endpoint security for the general public by building a safer cyber world. The Canadian cybersecurity landscape is at risk, and Canada currently trains less than half of the skilled professionals needed in cybersecurity-related industrials. The proposed research will address this large professional shortage in Canada for both the public and private sectors. Training of HQP in this program will include demanding skills in binary analysis, vulnerability analysis, reverse engineering, large-scale data analysis, and explainable machine learning. I expect that three PhD students, six MSc students, and five undergraduate students will receive training through this program that will prepare them to launch careers in academia, industry or government agencies.

随着超高速5G网络的引入，公司和组织基础设施面临着越来越多的终端设备，如PC、移动设备、物联网(IoT)传感器和执行器。它们显著扩大了攻击面，网络攻击已从外围设备极大地转移到终端设备。恶意软件感染和无文件漏洞利用是针对终端的两大快速发展的威胁，给公司和组织造成了重大经济损失。由于其集成简单性、低开销和对抗未来未知攻击的有效性，Deep Learning(DL)一直在推动下一代恶意软件检测和漏洞发现解决方案，以优化业界和学术界的终端安全。然而，作为黑盒方法的这些基于数字图书馆的解决方案无法向安全分析师和官员提供与传统的基于签名的解决方案相同程度的洞察力和可操作信息。 拟议的研究解决了针对终端设备上新出现的恶意软件和漏洞利用的黑盒DL解决方案的有效性、可解释性和可操作性等关键而紧迫的问题。该研究研究了软件生态系统中恶意软件和漏洞的潜在通用表示及其关系，以构建有效的DL系统来防御传入的攻击并从中获得洞察力。这项研究将是促进加拿大网络能力的基础，以发现、防御和应对针对公共和私营部门的新兴大规模网络攻击。它还通过构建一个更安全的网络世界来增强普通公众的终端安全。 加拿大的网络安全形势岌岌可危，加拿大目前培训的网络安全相关行业所需的熟练专业人员不到一半。拟议的研究将解决加拿大公共和私营部门的大量专业人员短缺问题。HQP在这个项目中的培训将包括要求掌握的二进制分析、漏洞分析、逆向工程、大规模数据分析和可解释机器学习方面的技能。我预计三名博士生、六名硕士学生和五名本科生将通过这个项目接受培训，为他们在学术界、行业或政府机构开始职业生涯做好准备。