On the surface our search histories do not appear to be particularly sensitive information and they are unlikely to reveal our identity. However, recent history demonstrates the flaws in this logic and show that our search histories can easily reveal our identity. In August 2006 AOL released 20 million "anonymized" search queries from approximately 650,000 users to the research community.
AOL anonymized these search histories by obfuscating or removing 'personally identifiable information' such as usernames and IP addresses. AOL replaced usernames with randomized unique identifiers.
Reporters from the New York Times analyzed these search histories and were able to
identify user #4417749. The reporters noted that user #4417749 searched for
- landscapers in Lilburn, Ga
- 60 single men
- homes sold in shadow lake subdivision gwinnett county georgia
- several people with the last name Arnold
A quick reference of other outside sources lead the Times reporters to Thelma Arnold, a now 65-year widow living in Lilburn, Georgia.
This example reveals the false promise of anonymization. In particular, it is difficult for a database administrator to anonymize one data source when the administrator does not know what other data sources an investigator is able to access. In this case, the Times reporters were able to use the phone book or perhaps property records from Lilburn, Georgia. The combination of these data sources allowed the reporters to identify Thelma Arnold.