Rakuten Securities said a third party broke into its systems and may have exposed data on up to about 11.25 million accounts, some of it held on a test-environment system where experts say customer records do not belong.
The brokerage detected the unauthorized access on November 16 and cut off access to the affected systems. A later investigation found that some information managed inside a test-environment system may have leaked externally on November 11.
What may have leaked
The exposed data covers up to about 11.25 million accounts: branch codes that identify offices, customer codes and login IDs. For up to about 4.47 million accounts, names, dates of birth, addresses, phone numbers, email addresses, employers and withdrawal account details may also have leaked.
Self-declared annual income and financial assets, asset holdings by product and last login dates are also in the possible leak. The data includes former customers who have closed their accounts. Login IDs differ from Rakuten IDs.
Why real data sat on a test server
Hiroaki Kuramochi, a director and senior executive officer serving as CTO of the security firm LAC, said keeping personal data in a test environment is not standard practice. “Using test data is the fundamental rule,” he said. “Even if production data is reused, proper masking is the rule.”
Rakuten Securities told ITmedia News why production data was there. “With the upgrade of our internal systems, some functions required checks using production-environment data,” the company said. It had copied part of the customer information held in production to a test server for a set period and run checks on it.
The company called this an exceptional step. Normally only dummy data sits on test environments and servers, it said, and production information is not placed there. It also said a single employee did not act on his own authority and that formal internal procedures were followed.
Kuramochi said building exhaustive test patterns is hard work. With customers on many different registration patterns, verifying added functions and reproducing faults requires checking multiple test cases, he said, which may explain the wish to use production data. “But creating dummy test data, or masking as I mentioned, is the fundamental rule,” he said.
Asked whether the same practice had been used before, Rakuten Securities told ITmedia News it may not have been zero but was close to it, and that this case was unusual.
