AI and Security Problems

This post covers some interesting problem areas at the intersection of AI and Security. These could be worked on in an hackathon format.

Foundational AI and Data Analysis

  • Learn about nitty gritty of AI with in-depth LLM101 from Andrej Karpathy and create your own version of AI Story Teller or a Security Game!
  • Solve popular problems from Kaggle like Titanic Bias Analysis. Answers questions like what is AI good for?

Science of AI, and AI of Science

  • ** Deduce Physical Laws with AI ** Use pySR to learn and find popular symbolic equations. Try to tackle the problem of can we create science with AI?

Enhancing foundation models with RAGs and Finetuning

  • ** Use an existing LLM model to optimize it for security use case ** Use one of the open weights models from huggingface to create finetuned model for security or enable retrieval augmented generation (RAG) on it. You can do RAG on generative model with APIs as well.

AI Devices enhancing Security

  • ** Use ESP32S3 to create a voice bot ** Make an voice bot inspired by this.
  • ** Use ESP32S3 to create a camera detection system with ChatGPT ** Continuosly take photos of a server enclosure area and detect if there was an intrusion. You can use inspiration and code from [friend][https://github.com/BasedHardware/Friend)
  • ** Deploy models on ESP32S3 ** Make your own AI system to detect an object (like recycleable or not) and train it. Samples are available on seedstudio

Generative AI Applications in Cyber Security

  • AI-Powered Intrusion Detection System Develop an AI-assisted intrusion detection system that can analyze network traffic patterns and system logs to identify potential security breaches in real-time.

  • Smart IoT Security Camera for Server Racks Create an AI-powered IoT security camera system that monitors server racks, detects unauthorized access or equipment removal, and sends instant alerts to security personnel.

  • AI-Enhanced Red Team Automation Design an AI system that can automate various aspects of red team operations, including vulnerability scanning, exploitation, and reporting, to improve the efficiency of penetration testing.

  • Automated Security Code Review Develop an AI tool that can automatically review code for security vulnerabilities, suggesting fixes and best practices to developers in real-time.

  • AI-Driven Project Planning and Threat Modeling Create an AI assistant that can help project managers and security professionals in identifying potential threats, conducting risk assessments, and developing comprehensive security plans for software projects.

Advanced Security Applications with Generative AI.

  • AI-Enhanced pfSense Firewall Integrate AI capabilities into the open-source pfSense firewall to improve threat detection, automate rule creation, and enhance overall network security.

  • AI-Powered Phishing Detection and Prevention Develop an AI system that can analyze emails, links, and attachments to identify and prevent sophisticated phishing attempts in real-time.

  • Automated Incident Response Orchestration Create an AI-driven incident response system that can automatically detect security incidents, prioritize them, and orchestrate appropriate response actions across multiple security tools.

  • AI-Based Malware Behavior Analysis Design an AI system that can analyze the behavior of unknown files and processes to detect and classify new or evolving malware threats.

Security of AI

  • Prompt Engineering
  • Secure AI Model Development and Deployment Develop a framework and set of tools to ensure the secure development, testing, and deployment of AI models, protecting against adversarial attacks and data poisoning.

What’s next?

To assist with the Merritt Secure AI Hackathon we will be doing a pre-hackathon introduction to AI and Security webinar on Friday July 19th from 3-4pm. This webinar will also be recorded.

Use the following zoom link to join on Friday, July 19th 3-4pm: https://peralta-edu.zoom.us/j/9402327279 Topic: Getting Started with AI and Cyber Security

Update: Recording is available here

References