AI-assisted User-friendly Threat Modeling : AI-assisted Features and Intelligent User Interface Design that Support Users and Improve System Usability in Threat Modeling Output

Sammanfattning

As a pillar of cybersecurity, threat modeling is becoming increasingly critical across industries as a technique to maintain proactive system security, involving various users in its workflow. However, existing threat modeling systems often suffer from significant usability challenges, especially when non-expert users are involved. To tackle these challenges, recent studies have introduced AI support and the concept of Intelligent User Interface (IUI) to improve threat modeling usability. Yet, the output modules of threat modeling systems, where users interpret identified threats and assign appropriate mitigations, remain under-supported and hard to use. Utilizing Research through Design, this study explores how to design meaningful AI features for user support and their implementation in Intelligent User Interfaces, specifically in the context of threat modeling output. Through qualitative inquiry, user-centered design, and empirical evaluation, this research identifies and proposes design requirements for AI and IUI features that better support users in threat modeling output module. Results show that users benefit from AI features such as proactive guidance, extensive explanation, and prioritized user control. These benefits are especially evident in hybrid IUI designs that combine traditional threat modeling output interface and AI assistant chatbot. The findings further illustrate the tension between efficiency and learnability in threat modeling system IUI design, suggesting the need for adaptive intelligent interfaces that are customizable based on user expertise and preferences. The study was conducted in collaboration with ABB Corporate Research, Sweden.