# Iris Ohyama, security breach involving 7.7 million records

> Iris Ohyama puts the number of people affected by unauthorized access to a third-party cloud service used by its online store at up to 7.7 million.
> By José Maldonado · Cybersecurity · 2026-11-09

Iris Ohyama makes plastic storage boxes, kitchen utensils and small household appliances, and in Japan it sells much of that catalog through Iris Plaza, its online store. The incident the company acknowledged in late October 2026 did not come through the store's servers: the unauthorized access occurred in a third-party cloud service the store used. That difference places the origin of the problem outside the perimeter the company controls.

The provisional balance that the company published on November 5, 2026 reaches 7.7 million people, according to the count detailed in [Iris Ohyama's notice](http://www.irisohayama.co.jp/news/2026/?date=1029)There are 5.8 million registered and former members, plus 1.9 million customers who bought without opening an account. The affected service held contact details and passwords, with differences depending on each person. The store and its related systems had been closed since October 28, a day before the first notice.

What matters is not whether the password was encrypted, but what can be built from the rest. A database with name, address, phone number and date of birth makes it possible to call someone by name, cite their last order and ask them to confirm a card the company never had. That is the fraud that comes later, and it does not require decrypting anything.

## Access to third-party cloud

The timeline published up to November 9, 2026, has three dates and no technical explanation: detection in late October, closure of the store on the 28th and the first notice on the 29th. The rest is what the company says it still does not know.

The unauthorized access was detected in late October in an external cloud service that the store used for part of its operations. The company suspended Iris Plaza on October 28 and, the following day, referred the case to Japan's Personal Information Protection Commission and to the police. The November 5 update added the detail that explains the scope of the incident: a [suspicious file within that third-party cloud](https://www.bleepingcomputer.com/news/security/iris-ohyama-data-cloud-breach-exposes-info-of-77-million-customers/). As of November 9, there is no record of direct communication from the attacker or of an identified perpetrator.

## The exposed data and the lesser of the problems

The inventory that the company acknowledges as possibly exposed varies from person to person, and that variation matters more than it seems: two affected customers can end up with very different risk profiles.

*   Full name
*   Postal address
*   Telephone
*   Email address
*   Gender
*   Date of birth
*   Password, only in the case of active members

Active members' passwords are stored encrypted, and the company has not detailed which scheme it uses. Encryption means there is a key capable of restoring the original text, so the right question is not whether the data was protected but where the key was and who could use it. A hash is not reversed: it is compared. If the company had used a salted hash, stealing the file would not yield usable passwords. The difference between those two worlds fits into one word of the statement.

The rest of the package makes it possible to build a deception that sustains itself: a name for the greeting, an address to cite the delivery, a phone number for the message and a date of birth for the verification excuse. None of that requires decrypting a password.

## How is the impact distributed between members and non-members?

The numbers separate two populations with different obligations for the company and different risks for those who suffer them. A member has a purchase history and a password; someone who bought without an account left their data and nothing more.

Group

People

What they left at the store

Registered members and former members

5.8 million

Contact data, purchase history and password, the latter only among active members

Customers who bought without an account

1.9 million

Contact data and delivery address

The bulk of the impact falls on the members' side, and not because of volume. A former member who closed their account years ago is still in the database with their name and address, with no password to rotate and no reason to log back into the store, so they will probably find out late or not at all. For the company that lengthens the notification: it has to track down people who are no longer active customers.

The 1.9 million people without an account are the group that fits worst into any reasonable retention policy. They bought once, left the bare minimum for the package to reach their home and now appear in an incident no one is going to call them about. There, no loyalty justifies the risk, only a database that grew while no one was watching it.

## Third parties again, the pattern most breaches repeat

The vector repeats with an insistence that is no longer surprising. Over the past two years, many of the incidents that have ended with customer data out in the open began at a supplier, a marketing tool or a cloud service the victim was renting. The [the Gnosis Pay breach](https://observatorioblockchain.com/en/cybersecurity/brecha-gnosis-pay-expone-datos-titulares-tarjetas/) put cardholders' data at risk without the company that was attacked being the one hosting that data. The company that was attacked had controls; the provider had a key, and the key opened the same thing.

Not all large breaches go through a provider. The [3AM ransomware attack](https://observatorioblockchain.com/en/cybersecurity/3am-ransomware-colegio-37-equipos-ayuda-ia/) against a British school got in with stolen backup credentials, with no third party involved. The difference lies in who was holding the key.

What makes that path convenient for the attacker is the distance. The provider sees requests from a legitimate application, the victim sees a service that works and nobody sees the whole picture until the file that should not be there appears. When it is discovered, the trail is split between two companies that share neither logs nor an incident timeline, and the customer receives a letter summarizing what both agreed to tell.

## A suspicious file on a server that is not yours

The Iris Ohyama story fits into one sentence that any e-commerce manager should read aloud: the data we lost was on a server that was not ours. There was no spectacular failure in the store, no broken encryption, no attacker inside the network for months. There was a suspicious file in a rented cloud and 7.7 million people on the balance sheet.

Compared with the [Capcom incident and its 11 million records](https://observatorioblockchain.com/en/cybersecurity/capcom-brecha-11-millones-registros/)the 7.7 million from Iris Ohyama are a smaller figure. The detail is worse: this includes addresses, phone numbers and dates of birth of people who are not even customers.
