Fugaku supercomputer breach: attackers hit a status dashboard, then tried stolen logins
The RIKEN Center for Computational Science, home of Japan's Fugaku supercomputer, has disclosed unauthorised access attempts following a data leak. Between 10 and 13 November, an unknown party attacked a web dashboard that displays operational status data for the machine.
The attack appears to have gone well beyond the dashboard. The intruders reached several servers holding data on authorised users and installed malware on them. Those servers stored names, email addresses and credentials used to access supercomputers and services not run by RIKEN. The attacker has tried to use those credentials, and RIKEN is working with the organisations that host the other systems to find out whether any attempts succeeded.
RIKEN says it does not yet know how the attack happened, or exactly what was accessed or stolen. It has said 4,222 registered Fugaku users are at risk, along with 375 users of its smaller Rikyu supercomputer. On 13 November someone tried to log on to Fugaku, and RIKEN quickly took the machine offline. Fugaku ranks ninth on the latest Top 500 list, and RIKEN's reporting suggests that running a workload on either system could cause significant damage.
The outage is already disrupting researchers who were scheduled to use the machine. So far RIKEN has found no evidence that users' research data has leaked, but it is not yet sure what information left its network or whether any of it describes third parties. A detailed investigation has been promised. The incident follows a string of serious cyberattacks on Japanese cloud providers, telcos, retailers and government agencies.
Why it matters
A low-profile status dashboard became a path to servers holding credentials, which breaks the assumption that a read-only public page carries little risk. Because the stolen logins also work on systems run by other organisations, one breach can spread to partners, so anyone who shares credentials or access across organisations is exposed. It forces a decision about segmenting internet-facing pages and about how quickly credentials can be revoked once a leak is suspected.
What to do
- Review which public-facing dashboards or status pages can reach internal servers, and segment them from systems holding user data.
- Reset credentials that were stored on any server that may have been touched, including those used on third-party services.
- Check user-data servers for unexpected malware or software installs, and review logs for failed or unusual login attempts.
- Tell partner organisations quickly if shared credentials may be exposed so they can investigate.