The Application and Threat Intelligence (ATI) Department sits right at the center of cybersecurity development at Keysight, delivering cutting-edge threat intelligence across multiple security products.

This project is a new initiative to enter the rapidly evolving landscape of Agentic AI security. Analysts at Gartner predicts that by 2028 around 40% of enterprise apps will include agentic AI, but many projects may be vulnerable, as best practices for securing AI agents are still emerging. You'll help build a validation framework that puts AI security proxies to the test: a system designed to actively probe, stress and audit AI defenses, so we can confidently say whether they hold up against real attacks. This efficacy testing will be needed to certify next-generation AI security solutions.

This tool will act as an adversarial "Evaluator" to verify:

🔹 Bypass Attacks: using obfuscation to hide malicious commands from proxy filters. 🔹 Discovery Integrity: checking whether unauthorized or altered tool calls go unnoticed. 🔹 Stateful Security: seeing if the proxy correctly tracks context over time across multi-turn interactions. 🔹 Rug-Pull & Behavior Drift: testing if the proxy detects tools that dynamically change their logic or permissions mid-session to bypass initial security checks.

If you’re curious about cybersecurity, software engineering, and AI-powered automation, this is your chance to learn, contribute, and help shape the future of automated security. Join us and explore how AI can transform cybersecurity!

What you will gain:

• How agentic AI solutions work and how to test fine-grained security solutions for them.
• Create scripts that simulate complex, multi-stage cyber attacks against agentic AI systems.
• Quantify the performance vs. security trade-offs of AI firewalls.

If you know Python and basic Linux and Windows commands, we’ll teach you everything. AI & security knowledge is good to have – But if you’re eager to learn, we’ll guide you!

Skills required: Mandatory: python, basic linux, basic windows Good to have: AI/ML knowledge, Security knowledge, OpenAI