Fugaku: follow exposed keys to every external service
A key stored in a research environment can provide access to other organizations. Organize recovery around each trust relationship, with verification at every destination and criteria for restarting automation.
Читать на португальскомBy Pentesty
Start with the trust relationship
TechRadar’s November 21, 2026 report prompted this analysis. In its second official notice, dated November 20, RIKEN confirmed the exfiltration of credentials for accessing external systems, reporting 430 instances, and requested revocation of the corresponding keys at each destination. Some external connections were identified; this does not establish a widespread compromise of the reachable services.
For a business sharing data or computing capacity with a partner, the practical question is where that authorization ends. Describe the approved workflow: which process uses the key, under which account and for what work. Include the owner of the destination service. Recovery needs to address that relationship even when the machine storing the file belongs to another organization.
Build a map that can be closed out
Choose a set of integrations to begin with and agree on scope with their owners. For each identified key, record a fingerprint, the associated identity, authorized destinations and the automation that depends on it. Avoid copying the secret into spreadsheets or tickets. Mark any destination whose owner or authorization mechanism remains unconfirmed as an open item.
The map should also cover operational dependencies: file collection, result delivery, repositories and scheduled jobs. Agree with each team on the replacement order and validation window. The application secrets analysis helps examine how local access can connect to other services, without assuming that all of them were compromised.
Turn replacement into a bounded test
Agree on a test using synthetic data and a low-impact operation at the authorized destination. Define the expected result beforehand: access using the replacement key performs only the approved task, while the old authorization stops working. Prefer audit records and administrator confirmations to explain the outcome. Any attempt using the old key requires explicit scope and permission.
Observe the automated process after the change as well. A job may still use another account, a copy of the key or an authentication path missing from the initial inventory. Record the identity actually used and investigate discrepancies with the system owner. These proposed checks should produce clear open items without claiming that a small verification covers the entire environment.
Restart work with a decision record
For each integration, bring together removal of the old authorization, installation of the replacement, the agreed test and the limitations encountered. Keep local service recovery and external access review distinct. Restoring a processing queue or rebuilding a server should not automatically close map entries that belong to other administrators.
Present remaining dependencies and their effect on the work being restarted to the operations owner. Decide who accepts each exception and when it will be reviewed. The final record should explain why an integration resumed, which authorizations were replaced and what evidence is still missing. Use it as the starting point for the next review when the application or partner changes.
Frequently asked questions
Is deleting the key from its original server enough?
Include the authorization maintained at every destination in acceptance. Identify its administrator and record removal or replacement, alongside recovery of the original server.
Can a key with a passphrase leave the response queue?
Assess file protection together with authorized destinations and uses. Do not close the review on that characteristic alone; document the decision, exceptions and their owners.
How can a team test without affecting another organization’s work?
Agree on scope and permission with the destination administrator, use synthetic data and limit the operation. Prefer administrative evidence when an access attempt is not authorized.
What evidence helps approve an integration’s restart?
A record connecting identity, destination, removal of old authorization, replacement and the agreed test. Include limitations and the owner’s decision about remaining open items.
Sources and references
- TechRadar: Fugaku report dated November 21, 2026
- RIKEN: second official notice, published November 20, 2026
Original Pentesty analysis, published November 21, 2026. The proposed scenarios are editorial guidance, with scope and validation to be defined for each environment.