Latest

Cybersecurity

Onamae Desktop Cloud hack exposes more than 26,000 servers

GMO Internet confirms unauthorized access to Onamae Desktop Cloud, shuts down all plans and admits trading credentials may have been viewed

Onamae Desktop Cloud hack exposes more than 26,000 servers
Onamae Desktop Cloud hack exposes more than 26,000 servers

A customer inquiry ended with thousands of servers shut down in Japan. GMO Internet confirmed that a third party had installed foreign programs in customer environments of Onamae Desktop Cloud, the virtual server service it sells to keep trading bots running without interruption. Its first notice put the number of potentially affected machines at 26,739, a figure that measures possible exposure, not confirmed infections.

The shutdown began on October 21 at 04:00 local time and covered all plans of the service, which still has no reopening date. Behind Onamae.com is GMO Internet, one of Japan's largest domain registrars, which also sells web hosting and DNS. The compromised product was no ordinary hosting service: the company itself advertised it as the ideal option for keeping MetaTrader 4 running continuously, and that was the problem.

That was also where the credentials were. The company's second notice, published on October 22, admits that the logins stored on those machines, including those for the MT4 and MT5 tools and for trading accounts, may have been seen by the intruder, and that customers are already reporting trades they do not recognize. Card data was not on those servers, a partial relief for anyone who entrusted their money to a robot.

A simple external program and the chaos it causes

The sequence fits into four dates. On the 19th came the customer inquiry, on the 20th the foreign program was confirmed and new sign-ups were cut off along with the control panel, and on the 21st the shutdown began at 04:00 local time.

GMO Internet has explained that the investigation began when it received a customer inquiry, and that on October 22 it went to the police. An external cybersecurity firm is now reviewing the path of the intrusion, and the company has said it will share details as it confirms them. What is already settled is the timeline: two days passed from the initial inquiry to the shutdown.

What the company has not published is which program it was or how it got in. Neither the first notice on October 21 nor the second identified the malware family or the entry vector. Nor is there a count of machines with verified infection: the October 23 update describes the 26,739 as "servers that could be affected," not as machines with a program already installed.

With that data in hand, no customer can yet know whether their machine was compromised or merely exposed. That distinction determines whether the notice describes an incident that occurred or a risk still to be confirmed, and the doubt will last longer than the shutdown.

Trading credentials at the center of the incident

The company's second note, published on October 22, changed the tone of the message: it acknowledges that the credentials for the trading tools, MT4 and MT5 included, and those for the trading accounts stored on those servers may have been exposed. It also confirms that it has received reports of unauthorized trades in customer accounts and that it has shared the situation with the affected firms.

MetaTrader 4 and MetaTrader 5 are the platforms with which much of retail forex trading places orders, and leaving them on a remote server allows the strategy to keep working with the customer's computer turned off. That convenience concentrates two secrets on the same machine: the platform password and the broker account password. Whoever compromises one gets both if the user reused the password or left it written in the same place.

The most significant detail is that this server acted as a safe. Anyone who left the broker password there so as not to retype it every morning handed over the key along with the machine, and the intruder needed nothing more to trade.

The company's request is specific and has a condition many overlook. The password change must be made from a machine other than the affected server, and then you must log out to cut any open connection. Add to that reviewing history and balance, enabling two-factor authentication if the broker offers it, and rotating any credential stored on the machine.

What to check if you had services on those servers?

Anyone who used Desktop Cloud for more than a forex bot faces the same problem on another scale, because every secret that lived on those machines must be assumed to have been read. Exchange API keys and access tokens for a control panel shared a disk with the intruder, and rotating what was stored there is the only sensible response while there is no forensic analysis detailing what was taken.

  • Change passwords from another machine and log out of the server before reconnecting to it.
  • Review the trade history and balance of every account whose credentials were on the machine.
  • Enable two-factor authentication at the broker and on any panel that allows it, even if the provider does not require it.
  • Be wary of emails and calls that lean on this incident, because the company warns that it never asks for passwords or card data through those channels.

The shutdown works in customers' favor on one point and against them on another. While the machines remain off, no one can log in with stolen credentials to erase traces, but the legitimate owner who needs to review logs and backups is also left without access. GMO has not given a reopening date and has said it will announce when it verifies that the infrastructure is secure.

Who is accountable when the provider is the one compromised?

A virtual server is a delegation contract: someone else patches the hypervisor, someone else monitors the network and someone else responds when something gets in where it shouldn't. When the failure occurs on that side, the customer discovers that their control over the environment begins exactly where the provider's ends, and that the credentials they stored there no longer depend on their own hygiene.

The track record of the technology supply chain leaves little doubt about the pattern. The failures that prove costly are rarely at the user's end, but rather in the layer everyone takes for granted, as shown by the chain of failures surrounding Ledger amid leaks, implants and ecosystem vulnerabilities. The security of an account ends up being worth whatever the machine where it is typed is worth.

The alternative is not free either. Moving the service to your own machine shifts the problem to the customer, who then takes on what a third party with more resources previously managed to close out the incident.

Setting up the service at home or on a separately contracted server leaves updates and backups in the user's hands, and anyone setting up their own server in 2026 takes on that work with method and with time. Neither option eliminates the risk: it only decides who suffers it when it arrives.

A critical service with no reopening date

GMO has said that the return of the service will be announced once the security of the infrastructure is confirmed and that the impact on its consolidated results is still being assessed. For the customer, the timeline matters more than the statement, because every day of shutdown is a strategy halted, and anyone trading with real money needs to know whether their account was exposed before reconnecting.

The weak point in this story was not a careless user's password, but the machine where that password slept. Anyone setting up a service like this again would do well to ask what the provider stores, what it logs and how long it takes to notify, because next time the clock will start running on their trading account and not on the Onamae panel.

José Maldonado

Author

José Maldonado

Activist and blogger on technology, free software and blockchain.

The latest
From Huang to Botín, from Zara to Byron: corporate responsibility in global capitalism

Keep reading