Building an AI Agent for Understandable Security
Cyber Security
Client company:Zolder B.V. Ontwikkelingsuitdaging
Adnan AL-Habsi
Ahmed Awlaqi
Mihai Glodici
Dave Derosiers
Gehad Abass
Mahamed Yahye
Project description
What is the best approach to developing an AI agent that assists users of AtticSecurity.com in monitoring and responding to Office 365 security incidents while ensuring the solution is cost-effective?
Context
Situation - Zolder B.V. supports SMEs in managing cybersecurity. Many SMEs struggle with interpreting security incidents, especially in Office 365 environments.
Problem statement - The complexity of security tools leads to inefficiency and high dependency on IT support for SMEs.
Objective - Develop an AI agent to simplify incident interpretation and provide actionable insights for non-technical users.
Conclusion - This chapter defines the challenges and objectives, forming the basis for the research approach.
Results
An AI chat-bot using ollama to serve the models on a virtual machine hosted in net-lab that can respond to human questions with also pre-defined queries and a front-end design which was an additional feature.
Downloads
Compliance Document for AI Agent Deployment (PDF)
Research Locking Down AI (.docx)
Zolder AI agent project plan (.docx)
About the project group
- Weekly meetings (Wednesday) with the Coach are conducted at TQ.
- Per sprint meeting with the product owner
- Project updates on documentation and progress will also be sent to MS Teams → Files
- During the meeting, the team will provide an update on their progress. They will explain what they managed to accomplish during the previous week and will ask further questions if needed.
- If a team member happens to be absent on meeting day, he should send a message to the others, notifying them in time, so there is no confusion between them