|
Amidst the sea of security announcements slated for the next week is a card swipe device that claims almost instant encryption of cards, avoiding the problem of card data being grabbed before encryption. Such claims are commonplace, but the VeriShield Protect from Verifone is making claims that—if ultimately proven true—would significantly advance retail payment security. The new unit uses Hidden Triple Data Encryption Standard (H-TDES) from a company called Semtek Innovation Solutions Corp.. It’s hardware unit is designed to deactivate if anyone succeeds in opening the case, making the planting of physical data-capture devices more challenging. Read more. |
April 4th, 2008 at 10:50 pm
After reading the Semtek chief’’s statement in your article, especially the statements regarding “Rijndael [AES] or Feistel techniques”, I’m not personally convinced that they haven’t rolled their own algorithm. What he described sounds “more complex than it should be” which is a definitely a flag. If they tried to write their own low-level encryption routine instead of using ordinary 3DES, it’s possible they unwittingly compromised the security of the standard encryption.
One of the modes of operation applicable to any block cipher is called output feedback mode (OFB http://en.wikipedia.org/wiki/Block_cipher_modes_of_operation#Output_feedback_.28OFB.29 ), which long ago solved the problem of encrypting variable data with a fixed block size algorithm by turning it into a key generator for a stream cipher. (So long ago that its patent has already expired.) Why would he claim they had to use these Feistel or AES techniques when there’s already an industry accepted solution?
There are ways to accomplish the task they describe using nothing more complex than 3DES and some simple transformations that mathematically retain the security of the underlying encryption.
To provide for backward compatibility, using 3DES in OFB mode the output could be masked down to the same size as the current character set size of the track, as long as the decrypting end also masked off the high bits. They theoretically could even leave the ISO-7813 sentinels, format codes, and parity characters unencrypted — even though that seems like a crib for a cryptanalyst, it shouldn’t come as a surprise to anyone that a magnetic card reader head is outputting encrypted track data.
There is even a fairly simple way to retain the PAN as digits, and the cardholder name as alphanumeric characters, while retaining the full security of the 3DES algorithm.
If Semtek is confident of their security, they should publish their algorithms and protocols so that they can be given a proper cryptographic review by qualified experts. (See Kerckhoff’s Principle http://en.wikipedia.org/wiki/Kerckhoffs%27_principle ) Until they do, their algorithm will carry a cloud of doubt.
April 7th, 2008 at 1:31 pm
Your last statement about proper cryptographic review by experts was the point of my quoted comment. Until the algorithms are proven secure and have test cases and numbers backing up how long a brute force attack would take to crack the data, this technology will have a hard time getting traction. On the other hand, if it can be proven secure by the cypher experts, then it will be a great addition to the PCI security toolbox.