Canadian ISP Hijacking DNS Lookup Errors
Freshly Exhumed tips us to news that Canadian ISP Rogers Cable appears to be redirecting invalid DNS requests to their own search and advertising page. Roadrunner got caught doing the same thing earlier this year. According to the article, "The hijacking appears to be an attempt by Rogers to use its Deep Packet Inspection (DPI) technology to cash in on the mistakes of its users." Freshly Exhumed also reminds us, "As IOActive security researcher Dan Kaminsky has warned in the past, this presents a very serious security problem."
I know one problem it can cause is for a number of spam tests which look for the message coming from a legitimate domain. When the DNS server says "yup, that resolves" even when there's actually no domain, the test is defeated.
The world's burning. Moped Jesus spotted on I50. Details at 11.
This must be brand new. I did a test just now and a bad URL sends you here:
http://www20.search.rogers.com/search?
With appropriate variables substituted for what you were typing of course, like this:
Enter: http://www.rogersblowz.com and you get:
http://www20.search.rogers.com/search?qo=www.rogersblowz.com&rn=mEelOh0JrKFZejZ
Let the debate rage on!!!
Mark
http://www.opendns.com/
basically it is remove your ISP's dns#s and add these
208.67.222.222
208.67.220.220
Politics is Treachery, Religion is Brainwashing
If the ISP is messing with the DNS service, the best thing to do is to use a different service.
For Linux/Unix users, you can just run a caching-only server on the desktop system, and it will issue its own name requests from the root on down. I've been doing a slightly more complex version of this at home for VPN purposes. (Forward requests to my employer's net to the private internal DNS server (through the VPN), while querying the public internet for all other servers.)
I don't know it a similar option is available for Windows users w/o shelling out big bucks, but it is technically feasible
If you cannot run a caching-only server, another option is to use a third-party DNS server. The only problem here is that it would not be automagically configured by DHCP, and would have to be manually set up.
This type of behavior is wrong on so many levels so I wonder what would be the danger of having ICANN police this type of behavior? It seems that ISPs are doing more and more to circumvent "standards" for their own gain. Would it be too much to ask ICANN to come up with a set of rules that ALL ISPs must adhere to or risk losing their netblock? I'm not even sure ICANN would do anything but I'm just posing the question.
Let me guess... They either already have, or soon will in a pitiful pretense of response to criticism, offer some sort of insanely weak opt-out mechanism.
I'm guessing one of two things:
Manually configure alternate DNS servers on a per device basis(a la Verizon's current setup, may they be thrice cursed)
or:
Something involving cookies, a la Phorm and friends.
For things like this, opt-out just isn't good enough.
[This is Dan Kaminsky]
I took a look at what Rogers is doing. They're using PaxFire, who indeed was directly vulnerable to the attacks I described at Toorcon a few months ago. PaxFire fixed their stuff up, but yes, the security of the web at Rogers is limited to the security of those ad servers at PaxFire.
I've had to do this, and it works. No annoying Verizon snatching my failed DNS lookups!
Of course, if you try to get this out of their so-called "tech support", they will not know what you're asking for until you manage to get down to tier 2 or 3 or so. Amazing as it sounds, teir-one Verizon Fios tech support will glaze over at the mere mention of DNS, and will stupidly keep trying to get you to do inane things with your browser.
Ruby Neural Evolution of Augmenting Topologies
I guess the thought with the ISP's nowadays is that "everybody else is doing it, why can't we?"
09 F9 11 02 9D 74 E3 5B D8 41 56 C5 63 56 88 C0 is the magic number.
It is not recommended to set immutable bit, as it causes issues in various situations (like restoring /etc from a backup).
Failure to write to an immutable file is an API issue unique to Linux boxes that use ext2fs or ext3fs.. Systems that run ReiserFS, XFS, or jfs, don't have this bug.
Future versions of DHCPD/Ifplug, or the C library, may very well properly detect the 'immutable' bit and clear it, before writing, then re-set the bit after finishing.
Just like they do if you're root and try to write to a file that exists with mode 444.
Essentially, immutable bit was historically a half-baked feature intended to be used with 'securelevel'.
The concept is you are able to mark important system files immutable, and then raise the securelevel. Once the securelevel is raised, the filesystem will not allow important system to be changed without booting in single user mode.
The removal of securelevel from the kernel in 2.4.x likely means that the days of the 'immutable' bit are numbered as well. Some day you may upgrade your kernel, and be surprised to find out immutable doesn't do anything anymore.
The reliable way to turn off gathering of DNS settings from DHCP is to use distro-specific instructions.
For example, in Redhat-based distros you edit /etc/sysconfig/network and specify "PEERDNS=no"
Of course, now that you understand the risk that the immutable bit may stop working for you unexpectedly later, you can go ahead and try setting it anyways... because it's easy, and simpler than configuring your network software the right way.
http://www.opendns.com?
Evolution is a state-sponsored, state-protected religion.
opendns.com does the very mangling I want to avoid and calls it a feature. At least they tell you they are doing it, and use it for stuff that could benefit end users (filtering allowed site names) as well as their own advertising. But it doesn't solve the problem. It is just a more "open" and up front version of the problem.