Collaborative Research: Education DCL: EAGER: Harnessing the Power of Large Language Models in Digital Forensics Education at MSI and HBCU

不断升级的网络犯罪威胁突出表明，迫切需要熟练的专业人员，精通为法律程序和商业决策收集和提出证据。然而，存储在设备、网络和社交媒体平台上的海量数字数据给定位和分析特定证据带来了挑战。连接证据和识别模式的任务对人类研究人员来说是一个艰巨的挑战。该项目的新颖性在于利用大型语言模型(LLM)的非凡能力为数字取证专业人员和学生创建定制的教育材料。这些材料旨在使调查人员掌握必要的知识和技能，以驾驭错综复杂的网络犯罪格局，并提高他们打击此类犯罪的效力。该项目的更广泛意义在于更好地为调查人员做好准备，以利用LLM辅助的技术进行数字取证，确保他们能够有效地适应网络威胁的演变性质。该项目将微调LLM，以根据广泛认可的储存库中的刑事案件构建数字法医调查图(DFIGs)。这些DFIG使用STIX(一种用于交换结构化威胁情报数据的标准化语言)作为调查过程、证据实体及其相互联系的可视化信息表示。为了确保准确性，图形中的实体和关系将通过图形神经网络(GNN)模型进行仔细检查，以识别和纠正潜在的错误。在包括课堂讲稿、案例研究和动手实验练习在内的全面教学材料的支持下，学生将被引导完成获取必要的专业知识的过程，以构建和分析针对不同数字取证案件的DFIG。这将促进为少数族裔服务的巴尔的摩大学、HBCU旗下的佛罗里达农工大学等机构的数字取证教育。此外，教师发展研讨会将向更广泛的国家社区传播教学材料，培养一个更强大、更具包容性的网络犯罪斗士网络。这一奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The escalating threat of cybercrime has underscored the urgent need for skilled professionals proficient in collecting and presenting evidence for legal proceedings and business decision-making. However, the vast volume of digital data stored across devices, networks, and social media platforms makes it challenging to locate and analyze specific pieces of evidence. The task of connecting evidence and identifying patterns presents a daunting challenge for human investigators. The novelty of this project lies in harnessing the extraordinary capabilities of Large Language Models (LLMs) to create tailored educational materials for digital forensics professionals and students. These materials are designed to equip investigators with the knowledge and skills necessary to navigate the intricate landscape of cybercrimes and enhance their effectiveness in combating such offenses. The project's broader significance is to better prepare investigators to leverage LLM-assisted techniques for digital forensics, ensuring they can adapt to the evolving nature of cyber threats effectively. The project will fine-tune an LLM to construct Digital Forensic Investigation Graphs (DFIGs) based on criminal cases from a widely recognized repository. These DFIGs serve as visually informative representations of the investigation process, evidence entities, and their interconnections using STIX, a standardized language for exchanging structured threat intelligence data. To ensure accuracy, the entities and relationships within the graphs will undergo scrutiny through graph neural network (GNN) models, identifying and rectifying potential errors. Supported by comprehensive instructional materials, including lecture notes, case studies, and hands-on lab exercises, students will be guided through the process of acquiring the necessary expertise to construct and analyze DFIGs for diverse digital forensic cases. This will promote digital forensics education at the University of Baltimore, a Minority-Serving Institution, and Florida A&M University, an HBCU, among others. Additionally, a faculty development workshop will disseminate the instructional materials to the broader national community, fostering a stronger and more inclusive network of cybercrime fighters.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.