RC RANDOM CHAOS

Firecracker Escape Chains Host Memory Access to Root

· via Hacker News

Original source

Firecracker VM Escape: PwnEscape Part 2

Hacker News →

Researchers demonstrated a complete escape from Firecracker microVMs, achieving root-level code execution on the host system. The exploit leverages an existing host memory read primitive to bootstrap a multi-stage attack that ultimately hijacks the thread managing the guest’s virtual CPU. The attack exploits how the Firecracker process interfaces with the KVM hypervisor through the KVM_RUN system call, where the kernel saves the thread’s register state.

The exploitation chain works through interconnected reconnaissance stages. Attackers place a known byte pattern in guest memory, then use Extended Page Tables (the virtualization layer’s address translation mechanism) to locate that page in host physical memory. From there, they reverse-walk the host’s page tables to identify the Firecracker process managing the guest. Using signatures in Firecracker’s binary code and calculating its load address to defeat ASLR, they locate the running instance in memory. The final step involves finding the saved CPU registers (pt_regs) belonging to the vCPU thread, which the kernel maintains during the KVM system call. By modifying these saved registers, the attacker can redirect execution when control returns from the hypervisor.

This demonstrates how multiple isolation mechanisms can be chained together when a single memory access primitive exists, transforming limited guest-side code execution into host-level compromise.

Read the full article

Continue reading at Hacker News →

This is an AI-generated summary. Read the original for the full story.