Mindgard – Cyber Security Platform for AI

Hacker News

Share card

Mindgard – Cyber Security Platform for AI

Hey Hacker News Community! Mindgard.ai (https://mindgard.ai/) is a way to assess, detect, and respond to cyber-attacks and data leakage against all forms of AI/ML, including LLMs, GenAI and any other AI assets including 3rd party supply chain. It also helps you to discover and learn more about the threats out there for AI. Our current platform version allows you to test different models and variations of attacks. We are Peter and Steve, the founders behind the Mindgard AI Security Labs at mindgard.ai. Our journey began with a simple yet challenging goal: to tackle the emerging cyber threats against AI and ML technologies all around the world. As a Professor in Computer Science at Lancaster University and expert in AI/ML systems – through R&D in a systems lab I foresaw the potential and need of the cyber security for AI. The outcome after years of research, preparation and work was the creation of a new technology capable of delivering unparalleled AI security testing, detection, and response. In May 2022, Mindgard was formed from the heart of a leading cyber security University lab within the UK. Mindgard AI Security Labs is not just another cyber security tool. It's a solution built from the ground up, with the help of top minds in the industry focusing on protecting your AI and ML models. At its core, Mindgard AI Security Labs is designed to help you understand cyber security threats against AI, learn about AI security, and how to conduct AI red teaming, and here's how it works: Through just a few button clicks, you can rapidly conduct AI security testing by designing, deploying and launching cyber attacks against different AI deployments for image, language, and LLMs. Upon submission, Mindgard AI Security Labs then deploys and launches the attack against the chosen AI deployed in Azure. Once completed, results from the AI security test are then collected, analysed and collated to generate interactive cyber risk reports of your AI deployment. If you’re interested in exploring how different attacks and AI models respond to different cyber attacks, In building Mindgard AI Security Labs, we have created from scratch a new technology that allows for automated and repeatable AI security testing and red teaming. Here are some things that it does: - Conducts AI red teaming and security testing in over 170 unique attack scenarios - Assess cyber risk of leading LLMs like Mistral - Demonstrates jailbreaking, data leakage, evasion, and model copying attacks - Easy selection of AI models, datasets and frameworks to be used in the AI attack scenario - Detailed reports on attack success rates Developing Mindgard AI Security Labs was not without its challenges. We spent years on research and development and gathering a large attack library that seemingly generalizes across different AI models and data, to really give an edge to our users. We're excited to share Mindgard AI Security Labs with the Hacker News community and are keen to hear your thoughts, suggestions, and questions. We believe Mindgard AI Security Labs can make a real difference in cyber security for AI, and we can't wait to see how you will use it to tackle secure adaption of AI throughout the company. Website: https://mindgard.ai/ Product: https://sandbox.mindgard.ai/