• The credit card clearing house EMC (who handles transactions for Visa, Mastercard and I think a few more credit card companies) decided not to business anymore with TellSell, the guys who'll send you 20 free knives if you order their set of stick-free pans. Reason: TellSell didn't sufficiently protect the credit card data of their customers, leading to too many fraudulent transactions. Apparently customer data were stolen via TellSell's site, and re-used on other sites by data identity thieves.
• The top story: the British government "lost" 25 million records of families entitled to child benefits. Each record would hold names and account numbers. It seems that the data were backed up or copied unto two CD's, and the CD's somehow got lost. It's not clear whether this case is a theft; the CD's may still show up (did they check their car trunks? It's where I usually find stuff that I've "lost").

In the case of the data exposure in Britain, I argue that there will always be f**kups due to human error. And even if we could eliminate the human factor: who's to guarantee that the technology of data protection is sufficient to prevent data exposure? It would obviously be best if the data were not in readable format anyway. Now I can see that the British "Ministry of Family Affairs" (there is no such thing, but bear with me) needs to know who's entitled to child benefits. They would need a list of families. But do they need the account numbers too? Why were those account numbers not stored separately, in a list maintained by the "Ministry of Finance"? The two lists might've been linked by keys based on non-reversible hash algorithms. If this had been the case, then the impact of the data exposure would've been far less: only a list of families would have leaked, without sensitive account data.

In the case of TellSell, I argue that this organisation doesn't need to keep credit card data at all. They only need a list of website visitors, and for each visitor, they need to construct a non-reversible hash value of credit card data. When appropriate, the hash value sent to EMC for payment purposes. Oddly enough, I wrote about using hash values for this purpose just two weeks ago.

In both cases, technological or human errors led to data exposure. But more fundamentally: the way that the data were stored was flawed when looking at the matter from a security perspective.

In most cases software will be designed and built so that it's delivered fast and cheap: meaning here: "store all data you can yourself, instead of delegating data ownership where it belongs." Delegating ownership would of course mean that complex algorithms of non-reversible hashes would need implementing, that transmission of data would involve transmitting such hashes and verifying them, and so on. Security-wise a good concept, but definitely more expensive than the current approach of least resistance.

I'm just wondering what will happen in the evaluation of these cases. I'm sure that the poor guy who backed up 25 million records onto two CD's and then misplaced them, will get some punishment. His boss will also suffer some consequences. But the IT architect who designed this? I'm pretty confident that he'll still be considered an expert who did a fine job. The same goes for TellSell. They will rigorously review their website security features and probably axe a few scapegoats, but they will never question the overall design.