Findings from the $1M Vercel Sandbox challenge
By Steven Van ·
Researchers submitted 1,285 reports and earned $242,000 in bounties; Vercel says none showed access to customer data.
Security researchers submitted 1,285 reports during a two-week challenge to escape Vercel Sandbox, earning $242,000 in bounties. Launched in August, the challenge uncovered one Linux kernel zero-day vulnerability. Vercel says no reports showed anyone accessing customer data.
A flaw in the Linux networking stack
The most significant report arrived on day three. HackerOne researcher Paulos Yibelo found that specially crafted packets sent from a standard sandbox, without special access, could leak host memory, crash the host or execute code on it.
The memory leak was demonstrated against Vercel production sandboxes. Host memory appeared in packets sent over the internet, and the leak could continue while the guest kept sending packets.
Where the attacks were demonstrated
The researcher demonstrated the crash and code execution attacks on their own KVM and Firecracker machines matching the Vercel Sandbox configuration. Two packets could crash the host and shut down every sandbox running on it. Exploiting the same flaw also allowed commands to run with full administrator privileges on the host, crossing the sandbox’s isolation boundary.
Security review and response
Vercel also worked with Trail of Bits to review the source code and security architecture. After receiving the kernel report, Vercel says it collaborated with the researcher to understand and mitigate the vulnerability, working with AWS and other industry partners to prioritize a fix.