Hi Ian, On Mon, Nov 19, 2012 at 5:24 AM, ianG <i...@iang.org> wrote: > On 19/11/12 18:19 PM, Jeffrey Walton wrote: >> >> An Adobe break in does not surprise me. >> >> Has anyone come across a paper on how to migrate an existing database >> with, for example, unsalted MD5 hashes, to something more appropriate >> for 2012? Naively, I don't see why MD5(password) cannot be an input to >> an improved system. That is, MD5(password) is just a pre-processing >> step to a system built with cryptographic legos. >> >> I'm trying to figure out why folks like Adobe (who know better and >> have the resources) are still using unsalted MD5. I suspect the answer >> has something to do with "its cost effective to be grossly negligent," >> but I want to give offenders the benefit of the doubt. > > Part of the issue is that in the world today, they aren't grossly negligent. > This has specific meanings, which are found in court. Until a court case > comes that finds such an act as grossly negligent, then they aren't. I don't disagree with you. I guess I'm putting the cart before the horse. But they are being grossly negligent (I chose that wisely I think). http://legal-dictionary.thefreedictionary.com/Gross+negligence.
> Which then speaks to the incentives. The corporations have internalised the > benefits of the model, and externalised the risks. Punters have to take the > costs of the risk of failure, so everyone on the supply side is happy. > Until a corporation has some skin in the risk game, they aren't going to do > more than punt all the risk to the consumer. Absolutely agreed. The risk analysis equations need to be unbalanced. As I said, its cost effective to be grossly negligent. I've participated in the risk acceptance process. Personally, I think risk acceptance is a broken process to accept defective software under an air of legitimacy. Nearly every problem that came across the team's desk was solvable by engineering - often times without re-architecture. To paraphrase you, privatize reward and democratize risk. > In which case, spending more when MD5 does a fine job is pointless; even if > it breaks there is less point in spending money. At the moment, when a > breach happens, companies are responsible for the direct internal losses, > short term reputation hit, and maybe the cost of a breach notification > service (which latter is either a bad joke that customers don't understand > or an insult). I'm a guy who was a victim of identity theft in the 1990s. I still have the stacks of letters trying to absolve myself. Two years of letter writing resulted in nothing, and it cost me nearly $10,000 to fix. Then it started again in the mid-2000s. I'm not fixing this one. And I did not even get the obligatory letter that cost less than $1 to mail (assuming it was due to a data breach back then). I can assure you corporations are not the only victims. > What's worse perhaps is that the first response of companies is to bolster > their defences with "best practices." This works well in court. "We do > what NIST said" and we're done and dusted. But as we know the risks are far > too complicated for some dry government inspired committee to navigate. In Adobe's case, they did not even do that ;) > We end up with a world in which there are companies that do real security > and risk work, and those that do "best practices." The latter group is far > larger, far louder, and unfortunately often more cost-effective. We (users and consumers) need legislative relief. Waiting for companies to "do the right thing" is ineffective. The proverbial snowball has a better chance surviving hell. I blame the legislature for that. Our representatives are bought, sold and traded like trading cards by special interest groups, lobbyist, and corporate america. Thank <diety> Osama Bin Laden did not make a PAC contribution on 9/10. History may have been re-written. One of the earliest democracies had it right - Sparta. They knew what class-A fuckups representatives are/were, and put them on trial when they left office. It was part of the process. Its funny how we've lost that lesson. A perfect case in point in former senator Dobbs, who took the illegal brides from Countrywide while directing US policy. When he got caught, he went on to lobby work and was quite blunt about PIPA (or whatever it was called): money for votes. I consider former senator Dobbs (and friends) more dangerous than Osama Bin Laden himself. I hope he (and friends) meet the same fate for what they have done to my democracy and my country. And I hope I am alive to watch it happen. > So, this incentives view will clearly change when a wave of class-action > suits declare companies grossly negligent (or some other legal theory). > And, they have to pay for it through external forces. We are starting to > see that now, with the first rulings coming out that find the banks & > suppliers responsible. But it is slow work - the legal cycle is at least as > slow as the cycle for systemic security improvements. I've had Google Alert years on "Data Breach" and "Class Action." The class has not gained any traction in penetrating and bringing change via the legal system. I think corporations realize this is a possible threat to profits, and its why they write in the no-enjoin clause in the Terms of Service and contracts. Fortunately, US courts have begun to bring some sanity to it http://www.topclassactions.com/lawsuit-settlements/lawsuit-news/2633-zapposcom-loses-arbitration-bid-in-data-breach-class-action-lawsuit. Sorry about the off-topic rant. Jeff >> http://www.h-online.com/security/news/item/Adobe-confirms-customer-data-breach-Update-1750344.html >> ... >> Update 15-11-12 14:55: According to security firm Sophos, the >> passwords were stored as unsalted MD5 hashes, which can easily be >> cracked quickly using modern CPU and GPU hardware. If the database >> extract turns out to be genuine, Adobe should have invested a little >> more effort in protecting the passwords of its users. _______________________________________________ cryptography mailing list cryptography@randombit.net http://lists.randombit.net/mailman/listinfo/cryptography
