ORCID

Abstract

Large language models (LLMS) have begun to demonstrate an increasing ca-pability to generate structured educational resources. Within the UK, second-ary schools face persistent challenges around the provision of suitable cyber security focused education. Ranging from lack of specialist teachers to une-ven access around high-quality teaching materials and limited curriculum technical depth. This paper investigates whether LLMs can effectively sup-port teachers by generating accurate, readable and curriculum-aligned cyber security learning materials for Key Stage 4 and Key Stage 5 students. Two experiments were conducted. The first examined how prompt design influ-ences the quality of generated resources, leading to the development of op-timized prompts for producing lesson plans, worksheets, homework tasks, and quizzes. The second evaluated these prompts across multiple models (GPT-4o, Claude Sonnet, and Gemini Flash) and across sixteen curriculum topics drawn from GCSE and A-level Computer Science specifications. Out-puts were evaluated using metrics measuring readability, curriculum cover-age, mistakes, and missing materials. The results indicate that LLMs can gen-erate teaching resources with strong curriculum coverage and low error rates. Most outputs achieved complete topic coverage, mistakes were infrequent and hallucinations were rare. Compared with existing freely available teach-ing resources, LLM-generated materials demonstrated improved coverage and comparable accuracy, although readability was occasionally reduced due to technical vocabulary. These findings suggest that LLMs can support cyber security education by providing adaptable teaching resources that teachers can refine for their specific classroom contexts. While teacher oversight re-mains essential, current LLMs already possess sufficient domain knowledge to support classroom use and may help improve the consistency and availa-bility of cyber security teaching materials.

Publication Date

2026-05-19

Publication Title

Human Aspects of Information Security and Assurance: 20th IFIP WG 11.12 International Symposium, HAISA 2026

Acceptance Date

2026-05-19

Keywords

AI in Education, AI, LLM, large language models (LLMs), cyber security education, secondary education

Download

Share

COinS