Work

Integrating Generative AI into Cybersecurity Education: A Study of OCR and Multimodal LLM-Assisted Instruction

Year: 2025

Type: conference-paper

Abstract: This research category full paper describes an LLM assisted instruction integration with a virtual cybersecurity lab platform. The rapid digital transformation of Fourth Industrial Revolution (4IR) systems is transforming today's workforce needs, increasing skill set gaps, especially for the older workforce. With an increasing emphasis on STEM skill sets like robotics, automation, artificial intelligence (AI), and security, the workforce will have to be re-skilled and up-skilled to meet future industry needs. Generative Artificial Intelligence (AI), has the potential to aid building of this workforce, by assisting trainee's by acting as instructional assistant, helping students build skill sets while performing experiential learning exercises. This paper presents a generative AI based instructional assistant integrated into a prior experiential learning platform. This instructional assistant uses a zero-shot, optical character recognition (OCR) based Large Language Model (LLM) assisted Instruction pipeline integrated into the legacy Cybersecurity Labs-as-a-Service (CLaaS) platform developed in 2015. The instructional agent assits the experiential learning by extracting the text from slide images using Tesseract OCR and to generate simplied instructions by prompting a general-purpose LLM enabling real-time instruction optimization with minimal infrastructure and integration effort. The instructional assistant was evaluated in a live university course and where in student feedback was collected using a 1-to-10 rating scale, with an average feedback score of 7.83 on 42 responses, indicating a strong perceived usefulness of the instructional assistant. In addition, we conducted a comparative study using multimodal LLMs that directly interpret slide images with the OCR pipeline, to analyze the computational overhead, and operational costs. While the multi-modal LLM pipeline performed better on slides with dense visual content, the OCR-LLM-based pipeline delivered comparable pedagogical value on text-centric slides with a significantly lower overhead, and operational costs. This work demonstrates that a lightweight and easily integrable pipeline can effectively bring modern generative AI capabilities into legacy platforms, offering a scalable enhancement for improving student comprehension in technical education settings.

Source: IEEE Frontiers in Education Conference (FIE)

DOI: 10.1109/fie63693.2025.11328318