Deanonymization

Companies regularly strip names and account numbers from data before sharing it with analysts, partners or the public. The hope is that the data stays useful while the people in it stay private.

Deanonymization shows that this hope can fail, because a few ordinary details can point to one person. The usual method is linkage.

Someone takes an anonymised dataset, such as card transactions showing a date, a merchant and an amount, and matches it against another source such as social media posts, loyalty records or public registers. Even a handful of data points, like a postcode, a birth year and a purchase pattern, can narrow a crowd of millions down to one individual.

In financial services the risk shows up in several places. Pseudonymous crypto wallets can be traced to real owners when they touch an exchange that collects identity documents, and shared credit data can be re-identified if the sample is small.

Regulators treat this as a privacy and security issue, with serious fines under data protection laws in many countries. For businesses, deanonymization is both a risk and a tool.

It is a risk when your own anonymised data is released and customers are exposed, and it is a tool when investigators use it to trace stolen funds or money laundering. Compliance teams must understand both sides.

Defences include removing rare details, grouping data into broader bands, adding statistical noise, and limiting who can see what. Regular testing by an independent team is also wise.

The techniques known as k-anonymity and differential privacy are used to measure and reduce the chance that any one person stands out. There is also a business and ethics angle.

Customers expect their financial data to be handled responsibly, and a re-identification incident can damage trust long after the technical problem is fixed, so many firms now run privacy reviews before releasing any dataset.