Cloudflare Fixes Container Isolation Flaw That Could Expose Customer Data
Cloudflare has fixed a container and sandbox vulnerability that could have allowed customers with Workers Paid accounts to recover residual data from other customers’ containers running on the same physical host.
Cloudflare Containers is available with Workers Paid plans and allows developers to run containerized applications on Cloudflare’s infrastructure alongside Cloudflare Workers.
Developers and companies use Cloudflare for backend services, processing jobs, and code execution environments.
Cloudflare Containers vulnerability could cross tenant boundaries
The flaw was reported through HackerOne on September 4 by Oren Yomtov, a security researcher at technology company Accomplish.
Exploiting the issue could have allowed an attacker to read files belonging to other customers. Potentially exposed information included directory listings, SQLite databases, Chromium profiles, .env files, and credential files.
According to Cloudflare’s disclosure, the issue occurred in a shared storage pool configured to skip zeroing reclaimed 64 KiB blocks.
“Once the thin volume backing up the container’s root disk was removed, its physical blocks were returned to a pool serving workloads belonging to multiple customer accounts,” Cloudflare explained.
The researchers were able to allocate 64 KiB of reclaimed physical blocks by writing only 4 KiB to unused space on a new container’s disk. Without the zeroing operation, the 4 KiB write would overwrite only part of the block, leaving the remaining 60 KiB available to be read. That residual data could have originated from containers previously using the same storage blocks.
Researchers found residual data in 18 of 24 container deployments and on 20 of 22 underlying nodes they tested. The recovered residue included directory structures, database pages, and structurally complete SQLite databases.
“This vulnerability could have allowed customers with Workers Paid accounts to recover residual data from storage blocks previously used by other customers’ containers on the same underlying host,” Cloudflare said.
“Successful exploitation could cross tenant isolation boundaries and leak file system metadata, directory structure, database pages, and application data.”
An attacker would not need control of the victim or the host to exploit the issue. However, the vulnerability did not allow attackers to read actively attached disks.
Cloudflare says customer data was not compromised
Cloudflare said its evaluation found no evidence that actual customer data was compromised in the manner described by Accomplish. The researchers used scripts that performed checks and returned aggregate numbers rather than collecting the actual contents of disk blocks.
The researchers also did not demonstrate that they could modify another customer’s data or disrupt workloads running on Cloudflare’s services.
Cloudflare removed the configuration that caused block zeroing to be skipped, decommissioned existing container disks, and cleared cached snapshots that could have contained stale mappings. The company completed all mitigation actions by September 19, 2026.
Cloudflare also reviewed logs, telemetry, and historical data as part of its investigation.
The fix was applied automatically to Cloudflare’s infrastructure, so customers do not need to take any action to address the risk.
Join Mikko Hypponen and security leaders from the NFL, Chanel, and Atlassian for a 2-hour digital summit about what will change with AI speed attacks, what defenders should stop doing, and how to verify, decide, fix, and revalidate at machine speed.
Source: www.bleepingcomputer.com