Since then I’ve been in correspondence with Netnames who argue pretty convincingly that the problem was not on their end. They say their records show no DNS changes for the period in question, and no one else they serve has has had a similar problem.
That means the DNS change happened at my hosting company (or, theoretically, somewhere else via cache poisoning, but that’s really unlikely). Hacking my DNS via my hosting company is certainly possible, even though they don’t manage the registration. That said, it’s sort off an odd thing to think happened because any hack capable of making that change there should have been able to do far far more damage and hit at least the other domains I manage from the same machine. Indeed, depending on the vector it could theoretically have hit all of them: my domains are spread over three machines; changing the DNS would require either root-like access on one of them or access to a control panel that gives some power over all of them. Yet none of those things happened. Maybe I dodged a bullet.
Meanwhile, I’ve changed all the relevant passwords (which were already strong random ones) and am working hard to plug every hole that my host’s automated security scan says it identified. Unfortunately, I have a lot of sites covering a very wide variety of personal and professional projects that have grown up over the years, and the scan resulted in a 12-page single-spaced list of things that might need fixing. It correctly identified some outdated installs of software packages, but the list of so-called hacked files seems overwhelmingly to consist of false positives (I’ve been investigating them with a simple text editor and so far they are mostly simple HTML files created by my cache program and fitted with legitimate headers) — so this is not an easy job.
Naturally, I wasn’t pleased, even if I sort of liked the middle part of the rap. Why attack me of all people? I’m for net freedom. Worse, the hack was blocking my main personal email address. Still worse, I was no longer able to access the domain via sFTP or ssh — everything timed out — making debugging somewhat challenging.
Eventually I figured out another way to log into the host machine, and verified that none of the files on the law.tm domain, including the .htaccess file, had been changed. This removed the most likely vector of the attack. That left two possibilities: The first was the very unlikely possibility of some very subtle SQL injection attack plus a level of traffic so hosing the domain that I couldn’t get through to it via ssh; this seemed unlikely because if there really was some DDOS-like event in progress I would have heard about it from my hosting company, Dreamhost, when the machine crashed, plus the redirect of the web page shouldn’t have worked either.
That left option #2 as the main suspect: a hack of the DNS records. The DNS records for this particular domain are manged by a different company than my web host, and their help desk is (a) located in London and (b) only open 9-5 London time Monday-Friday, leaving me high and dry for the weekend (smart hackers?). Perhaps coincidentally, perhaps not, I had just renewed the law.tm domain a few days earlier with the Netnames registrar. So the only thing I could do while I waited for the Netnames help desk to wake up was try to satisfy myself that this really was a DNS hack. That proved harder than I would have liked: the DNS records seemed to show incorrect information, with web requests for the domain being pointed to 18.104.22.168 and mail being sent to 22.214.171.124, neither of which was right. But then again sometimes the nslookup would come up ok with the right data. It could have been a propagation issue but why then were my http requests, even when I cleared DNS cache, never going through to my real page? Maybe, I worried, I didn’t know how to read the DNS records properly.
So I struggled with the problem. On Saturday I felt hamstrung by unusually slow and poor helpdesk support by Dreamhost, who have been much better in most of my past interactions. This time they announced a new-to-me policy that we couldn’t communicate by phone, only email, as they want a written record of anything relating to a security issue. And it took hours to get the first email response. When they did swing into action Dreamhost also refused to confirm I was having a DNS issue, even though that would have gotten them fully off the hook, saying only that the results were “ambiguous” … although in retrospect, that may have been an accurate assessment … so maybe score one for them after all. Unfortunately other than giving me an automated scan that showed possible problems elsewhere in things I manage but not on law.tm or its users. they didn’t say anything helpful about what else the problem might be.
In the end, thankfully, the problem seemed to solve itself this afternoon. The dig and nslookup data changed for the better — no more signs of the 126.96.36.199 or 188.8.131.52 IP numbers. OpenDNS’s cache started reporting the right info in more and more locations. Pretty soon all was back to normal. I even got a few — so far, sadly just a few — of the test messages I’d sent myself. (If you emailed me Friday evening or later, send it again please).
So I’m now pretty sure it was a DNS issue. Whether netnames got hacked (it’s happened before), or whether it’s some particularly ham-handed activity in connection with the domain name renewal, I may never know. Everyone I used to know at Netnames, which has been taken over once or twice since I last looked, seems long gone.
The NYT offers 512 Paths to the White House — a cute online app in which you choose how you think key swing states will come out and it tells you what other states the candidates have to get in order to win.
Give Romney Florida, and Obama Ohio, and then see just how many states Romney still needs to win. Basically if Obama takes Virginia OR Wisconsin plus any one of NC, Colorado, Iowa, Nevada, New Hampshire, he wins (except that New Hampshire + Wisconsin is a tie, which means the House will pick Romney).
A little morning fun, and a way to keep track as the results come in. Spotted via Talk Left.
Google has disabled use of the Maps API for this application. This site is not authorized to use the Google Maps client id provided. If you are the owner of this application, you can learn more about registering URLs here: https://developers.google.com/maps/documentation/business/guide#URLs
Coding error? NYT not paying its bill? The message popped up at the front page of a section, so it’s hard to see the NYT failing to have registered it.
One thing that always comes up when I discuss WikiLeaks with other lawyers, whatever country they may be from, is shared incredulity at Assange’s claim that he faces a great risk of hypothetical extradition to the US from Sweden, or that his risk would be greater there than in the UK…especially given that when it comes to extradition to the US, the UK has to be in the running for Top Poodle. But what do I know about Swedish law?
Thus, it it is reassuring and unsurprising to find a quality analysis of the Swedish legal rules relating to any hypothetical extradition request. And, no, according to Klamberg on Extraditing Assange from Sweden to the U.S. Assange wouldn’t seem to have much to worry about.
Which leads suspicious minds to wonder if perhaps there isn’t some other reason why Assange doesn’t want to be extradited to Sweden?
I thought I would post the slides from my P3P talk. I’m not sure if I will write this up into a paper. On the one hand, there’s really nothing surprising in what I’m saying here. On the other hand, there doesn’t seem to be a paper out there that directly addresses the topic, so there would at least be some point in writing it up.