One of the topics that prompted me to start posting again -- the recent announcement that a group of researchers had cracked the MD5 hash code.
But I'm a couple days late on the breaking news -- and i think this post by John Viega at ZDNet.com is one of the better ones at putting it into perspective: MD5/rogue CA attack: The sky is not falling
I don't think I can make his points beter, so I'll endorse his words instead.
So don't panic, OK? Nonetheless, it is quite serious, and illustrative of a couple points I want to make:
- It takes a long, long time to change things on the internet. We have spam because we have antiquated email protocols designed even before the term 'internet' was in use. They need to be changed, but it won't be easy. It's excruciatingly hard to get agreement on changes. (Not whether they're needed -- what to do instead). And it's hard to put changes in place, because you have to get millions of people to make the change.
- MD5 has been known to be vulnerable for quite a while, but CA's haven't changed. (Certificate Authorities -- the vendors that sell signed certificates that web servers use to prove to your browser that they are who they say they are). These are people who are supposed to be aware of security, so I'm scratching my head over that one. But there are a lot of people, myself included, who act as CAs for companies and other security uses besides browsers, and we need to stay on top of things like this as well. I hope we're not using MD5 -- but we might have been forced into it if some other piece of software wouldn't accept SHA-1.
- Security is hard. Actually, it's more than hard, it's impossible, in an absolute sense. So the risks are always there. As is so often the case in life, we always make a trade-off -- how much effort, inconvenience, cost, do we put out, to allleviate what risk?
- I don't agree with a lot of the risks that are being taken every day by vendors and implementors. Starting with choice of implementation techniques that are known to be problematic. A sociological puzzle. Perhaps more on that another day.
- I think a lot of individual risks to individual users are often overstated. An MD5 attack currently requires 200 game consoles, time, skill, and money, and another coordinated attack on the DNS infrastructure. But taken as a whole -- all risks, over all the population -- I think the overall impact is grossly understated. Among other things, I think there are major opportunity costs -- things we just don't even try to do because of the security environment.
