# How to get an ATM PIN in 15 guesses

## Bank employee fraud just got easier

Cambridge researchers have documented a worrying PIN cracking technique against the hardware security modules commonly used by bank ATMs.

Mike Bond and Piotr Zielinski have published a paper detailing how a complex mathematical attack can yield a PIN in an average of 15 guesses. By design, it shouldn't be possible to guess a four-digit pin in less than an average of 5,000 attempts.

The attack, documented in a paper published earlier this week, is directed against the decimalisation tables used to translate between a card PIN and the hexadecimal value of a PIN generated when the hardware security module checks the validity of a number.

The attack works not by going after the PIN number directly but by manipulating the contents of the decimalisation table in order to gain clues (such as which digits are or are not present in the PIN).

Refining the technique, which allows a PIN to determined in an average of 24 iterations, might allow an attack to succeed in 15 guesses. The methodology of the attack, too mathematically complex to be properly explained in the context of a news story, is explained here.

Mike Bond told us that the risk of attack comes from a corrupt insider, perhaps in computer operations and with access to sensitive manuals, who might be able to use the attack to refine what would otherwise be a brute force attempt to guess PIN numbers.

Fraud, in these circumstances, might still be possible. The attack is simply a more powerful, optimised means of cracking PIN numbers.

In their paper, Bond and Zielinski outline mechanisms banks might apply to guard against the attack.

In the short term, according to Bond, probably the best way to guard against the attack is to make sure it isn't possible to change the decimalisation table without permission. Longer term the researchers warn in their conclusions that "support for decimalisation is not a robust approach to PIN verification".

"Unskewed randomly generated PINs stored encrypted on an online database, as already used by some banks, are significantly more secure," Bond and Zielinski conclude.

As a stop gap an audit trail in ATM hardware security module will also allow the banks to spot when something suspicious occurs. This would allow banks to finger corrupt insider but it wouldn't necessarily protect customers, due to ongoing confusion in the UK's liability regulations for bank transactions.

Bond explained that UK case law does not as yet determine where liability lies in the case of disputed PIN-authorised transactions. With credit card purchases liability lies with merchants but in the question of liability has yet to be determined.

The wider consequences of the attack method documented by the Cambridge researchers once again spotlight this gap in UK law. ®