Predicting a person's social security number is a lot easier than previously thought, according to new scientific research that has important implications for identity theft.
Armed with publicly available information about where and when an individual was born, researchers from Carnegie Mellon University were able to guess the first five digits of a SSN on the first try for 44 percent of people born after 1989. The success rate balloons to as high as 90 percent for individuals born after 1989 in less populous states such as Vermont. Success rates also rise when the researchers got more guesses. The first five digits for six of 10 SSNs can be identified with just two attempts.
"We demonstrate that it is possible to predict, entirely from public data, narrow ranges of values wherein individual SSNs are likely to fall," the researchers wrote in a paper (PDF) published Monday in the Proceedings of the National Academy of Sciences journal. "Unless mitigating strategies are implemented, the predictability of SSNs expose them to risks of identify [sic] theft on mass scales."
A social security number is a nine-digit number issued to US citizens and other residents and that's used as a tax payer ID.
Researchers Alessandro Acquisti and Ralph Gross started with the hypothesis that SSNs should have easily detected patterns that correspond to the location and time of issuance. That's because the first five digits are based on the address on the application. The Social Security Administration publicly publishes the criteria by which these numbers are generated, ironically in an attempt to root out fraud by making it easier to detect phony numbers.
"The analysis confirmed the regularities we expected: As hypothesized, a strong correlation exists between dates of birth and all 9 SSN digits," the researchers concluded.
Using a publicly available database of SSNs that belonged to people who have died, the researchers were able on average to match on first try 7 percent of all records for individuals born nationwide between 1973 and 2003. That amounted to just 2 percent of SSNs for individuals born in California, the most populous state in the US. By contrast, they had a 90-percent success rate predicting the first five digits of people born in Vermont. The overall success rate rose to 60 percent when the researchers were given one or two attempts.
Once the first five digits have been identified, its not too hard to figure out the remaining four using standard brute-force techniques. The end result: It took 1,000 or fewer per target to match the entire nine-digit SSN for 0.8 percent of all records in the data base with dates of birth between 1973 and 1988. The success rate rose to 8.5 percent of all records with dates of birth between 1988 and 2003.
"When one or two attempts are sufficient to identify a large proportion of issued SSNs' first five digits, an attacker has incentives to invest resources into harvesting the remaining four from public documents or commercial services," they write. "More importantly, when <10, 100, or 1,000 attempts are sufficient to identify complete SSNs for massive amounts of targets, brute-force attacks replicating the algorithm we presented in the previous section become economically plausible."
And as the researchers point out, many services are only too happy to give up the last four digits of a person's SSN anyway.
The researchers plan to present their findings later this month at the Black Hat security conference in Las Vegas. ®