AI-assisted development environments are rapidly expanding the attack surface, and NomShub shows how quickly trusted workflows can turn into full system compromise. In this session, we break down a real-world exploit chain in the Cursor IDE, where a developer simply opens a repository and unknowingly triggers indirect prompt injection, a sandbox escape, and a persistent remote tunnel. The result is silent, durable shell access that blends into normal development activity and bypasses traditional security controls.
Join us to learn:
- How the NomShub attack chain moves from prompt injection to persistent shell access with no explicit user action
- How sandbox escape and remote tunneling in Cursor enable stealthy, durable access that evades EDR
- Practical detection and mitigation strategies for securing AI-driven development workflows
AI-assisted development environments are rapidly expanding the attack surface, and NomShub shows how quickly trusted workflows can turn into full system compromise. In this session, we break down a real-world exploit chain in the Cursor IDE, where a developer simply opens a repository and unknowingly triggers indirect prompt injection, a sandbox escape, and a persistent remote tunnel. The result is silent, durable shell access that blends into normal development activity and bypasses traditional security controls.
Join us to learn:
- How the NomShub attack chain moves from prompt injection to persistent shell access with no explicit user action
- How sandbox escape and remote tunneling in Cursor enable stealthy, durable access that evades EDR
- Practical detection and mitigation strategies for securing AI-driven development workflows
AI-assisted development environments are rapidly expanding the attack surface, and NomShub shows how quickly trusted workflows can turn into full system compromise. In this session, we break down a real-world exploit chain in the Cursor IDE, where a developer simply opens a repository and unknowingly triggers indirect prompt injection, a sandbox escape, and a persistent remote tunnel. The result is silent, durable shell access that blends into normal development activity and bypasses traditional security controls.
Join us to learn:
- How the NomShub attack chain moves from prompt injection to persistent shell access with no explicit user action
- How sandbox escape and remote tunneling in Cursor enable stealthy, durable access that evades EDR
- Practical detection and mitigation strategies for securing AI-driven development workflows
Speakers
