The domain-database is currently unavailable for i pulled the plug accidentally. I'm on my way to rebuild the db - and i used the opportunity to triple the server's ram. Maybe stuff gets faster this way.
tom
Wednesday, August 29, 2007
Monday, August 27, 2007
SEO with serversniff
Search-Engine-Spammers discovered that Serversniff.net is a lovely site where they can produce a page with links to their pages to be optimized. How does this work?
If you call a page like FileInfo you get an output with a link to your page or at least with a http://somedomain/file, that google will interpret as link to follow. The spammer do exactly this.
Serversniff uses google-adds to get at least a few bucks supporting the servercosts. Google-Bot does follow every customer, indexing the called page a few seconds later. The spamers use serverfarms or, more likely trojanized machines to call their page 5 times in a row from different machines all over the world. Crazy.
I dropped the links on some pages, and I will put on a noindex-Header in the metatags of all relevant Serversniff-pages, hoping to stop these totally useless crapsters from abusing my machine. Seems like nothing is too crazy on the internet.
tom
If you call a page like FileInfo you get an output with a link to your page or at least with a http://somedomain/file, that google will interpret as link to follow. The spammer do exactly this.
Serversniff uses google-adds to get at least a few bucks supporting the servercosts. Google-Bot does follow every customer, indexing the called page a few seconds later. The spamers use serverfarms or, more likely trojanized machines to call their page 5 times in a row from different machines all over the world. Crazy.
I dropped the links on some pages, and I will put on a noindex-Header in the metatags of all relevant Serversniff-pages, hoping to stop these totally useless crapsters from abusing my machine. Seems like nothing is too crazy on the internet.
tom
Sunday, August 19, 2007
I still don't like solaris.
I recently did a pentest. Whilst torturing a webapplication i came across a file-inclusion-vulnerability, allowing me a glance on /etc/passwd via opening http://tortured-site.xx/safe.php?file=/index.php.
I tried /etc/shadow, but no luck, apache was not running as root. I poked around to see that i was on a solaris-machine. F*ck.
I remembered that years ago, when i used to work as a webmaster, they gave me a sun E something. A horrible fast machine, but i had no clue from solaris. After two weeks I gave up with the awful thing and got a copy of the first beta-version of Suse-Linux for Sun. I don't like Suse either, but after installing Linux on it the Sun turned from a constant source of anger and stress into a horrible fast database-machine. I decided that solaris an me will never get friends.
So, what now? I knew i had to come up with something reasonable, not just a list of accounts to make the customers webmasters really aware of the problem. I tried to find other logs - but no way. Solaris' not Linux, and the Apache was homegrown installed in some funny directory i couldn't find.
Finally I came up with a promising solution: I had the accounts from /etc/password, and it was written there, that they all used /bin/bash. I went for /home//.bash_history. No luck on the first two accounts, but then, bingo, there it was: the admin deploying new software-versions, connecting to the database with user and password on the commandline and finally grepping through the apache-logs.
The webapp suffered from another minor vulnerability: All https-transactions were done via GET-Requests. I already pointed this out in my report that it's a really bad idea to transfer bank-account details and creditcard-numbers and cvcs in an URL, even if it is transferred via HTTPS.
From there on stuff was easy: I used the file-inclusion to download a days logfile, filtered the relevant requests via get and had lovely stuff to present:
What we can learn from this is:
The vulnerable safe.php proved to be a quick-and-dirty solution: 6 lines of custom code killing an otherwise really good webapp. Safe.php is fixed now, but hey, the site is huge, consists of many servers and I'm still keen on getting access to more user-data. Pentesting sometimes reminds me on an interactive version of mistery-stories and whodunits like "The three investigators" when i was young. I still like this part of my job. And, for i'm using zsh instead of bash, err, anybody can tell me how to disable .zsh_history on my machines?
tom
I tried /etc/shadow, but no luck, apache was not running as root. I poked around to see that i was on a solaris-machine. F*ck.
I remembered that years ago, when i used to work as a webmaster, they gave me a sun E something. A horrible fast machine, but i had no clue from solaris. After two weeks I gave up with the awful thing and got a copy of the first beta-version of Suse-Linux for Sun. I don't like Suse either, but after installing Linux on it the Sun turned from a constant source of anger and stress into a horrible fast database-machine. I decided that solaris an me will never get friends.
So, what now? I knew i had to come up with something reasonable, not just a list of accounts to make the customers webmasters really aware of the problem. I tried to find other logs - but no way. Solaris' not Linux, and the Apache was homegrown installed in some funny directory i couldn't find.
Finally I came up with a promising solution: I had the accounts from /etc/password, and it was written there, that they all used /bin/bash. I went for /home/
The webapp suffered from another minor vulnerability: All https-transactions were done via GET-Requests. I already pointed this out in my report that it's a really bad idea to transfer bank-account details and creditcard-numbers and cvcs in an URL, even if it is transferred via HTTPS.
From there on stuff was easy: I used the file-inclusion to download a days logfile, filtered the relevant requests via get and had lovely stuff to present:
- A list with thousands of CreditCards with CVC, Dates and Names.
- A nice record of what the various admins did over the last years. Some proved to be knowledgable and exact, even verified md5-sums of uploaded files to ensure their integrity, some proved to be unix-analphabets like me.
- A few passwords for accounts and the locations of ssh-keyfiles i didn't bother to download.
What we can learn from this is:
- It is a nice idea to make sure your webserver can't read anywhere on your partition outside the webroot.
- It is a nice idea to keep your bash_history-file small. Putting "export HISTFILESIZE=0" in your ~/.bashrc will do the trick.
- Consider all user input as dirty.
The vulnerable safe.php proved to be a quick-and-dirty solution: 6 lines of custom code killing an otherwise really good webapp. Safe.php is fixed now, but hey, the site is huge, consists of many servers and I'm still keen on getting access to more user-data. Pentesting sometimes reminds me on an interactive version of mistery-stories and whodunits like "The three investigators" when i was young. I still like this part of my job. And, for i'm using zsh instead of bash, err, anybody can tell me how to disable .zsh_history on my machines?
tom
Sunday, July 08, 2007
How nice
Usually people use serversniff. And they complain, when somethings completely not workin. Otherwise they don't give a shit. Really.
Ok, no real matter - serversniff is a hobby and the API is there to ease my daily work. Nothing more. But really, sometimes i wish people would care more.
Roelof Temmingh does care occasionally, for he uses parts of serversniff for his evolution. He contributed valuable code and (unknowingly) many many ideas and thoughts, and he's constantly begging for new functions. Roelof, if i had the time, i'd implement far more of your requests.
I had somebody asking for an API-Password recently - and a day after i had a bugreport. I was so glad that someone cared about the bugs that i fixed them right away. When I started with serversniff i had a dream of people bringing great ideas and great scripts helping me earn big $$$. Serversniff's live for around 2 years now - about 98% of the code is still mine, and the $$$ still don't pay for hardware, electricity and servercosts. Maybe i should stop ranting here.
Have a nice week,
tom
Ok, no real matter - serversniff is a hobby and the API is there to ease my daily work. Nothing more. But really, sometimes i wish people would care more.
Roelof Temmingh does care occasionally, for he uses parts of serversniff for his evolution. He contributed valuable code and (unknowingly) many many ideas and thoughts, and he's constantly begging for new functions. Roelof, if i had the time, i'd implement far more of your requests.
I had somebody asking for an API-Password recently - and a day after i had a bugreport. I was so glad that someone cared about the bugs that i fixed them right away. When I started with serversniff i had a dream of people bringing great ideas and great scripts helping me earn big $$$. Serversniff's live for around 2 years now - about 98% of the code is still mine, and the $$$ still don't pay for hardware, electricity and servercosts. Maybe i should stop ranting here.
Have a nice week,
tom
Friday, June 22, 2007
Statistics B
We're far from complete, but getting better:
The DB knows ns/mx-records for 12.829.155 domains. More is added daily until we're complete.
The DB knows currently 1.074.097 nameservers (counted by hostname, not ip!).
tom
The DB knows ns/mx-records for 12.829.155 domains. More is added daily until we're complete.
The DB knows currently 1.074.097 nameservers (counted by hostname, not ip!).
tom
Statistic-Figure A: 19.155.784
I'm not really into statistics and i don't do them regurlarly. But i'd like to remind myself that serversniff.net currently knows more than 19.155.784 unique domains with at least one resolving host. This should be around 15 percent of all known domains on the internet.
we're still adding around 100.000 new domainnames per day, focussing on domains outside of the .com/.net-space. To get a glimpse of domains added take a look at http://tomdns.net - there you can see the newest domains added in realtime.
tom
we're still adding around 100.000 new domainnames per day, focussing on domains outside of the .com/.net-space. To get a glimpse of domains added take a look at http://tomdns.net - there you can see the newest domains added in realtime.
tom
Spam from serversniff.net
Some asshole sent out a spam-wave with random Serversniff.net-Senderadresses. The little sucker put in a return-path and a sender with @serversniff.net. Since i have defined a "catch-all"-mailaccount for serversniff.net, i get all those nice complaints and returned mails. Hundreds of them! Argh.
And yes Sir, no M'am, neither my webserver nor my mailservers are hacked, take a look at the mailheaders:
Return-Path: <stasIsaev@serversniff.net>
Received: (qmail 25562 invoked by uid 0); 22 Jun 2007 12:17:06 +0300
Received: from 220.125.204.181 by post (envelope-from <stasIsaev@serversniff.net>, uid 92) with qmail-scanner-2.01
(clamdscan: 0.90/2659.
Clear:RC:0(220.125.204.181):.
Processed in 0.239443 secs); 22 Jun 2007 09:17:06 -0000
Received: from unknown (HELO ?220.125.204.181?) (220.125.204.181)
by post.ziniur.lt with SMTP; 22 Jun 2007 12:17:04 +0300
Received: from [220.125.204.181] (183.178.25.193)
by stasIsaev@serversniff.net with SMTP;
for <gvitkauskasd@ziniur.lt>; Fri, 22 Jun 2007 19:17:22 +0100
MIME-Version: 1.0
None of these sender-IPs belong to serversniff.net's infrastructure. Seems that it's time to drop the catch-all-adress for serversniff.net.
tom
And yes Sir, no M'am, neither my webserver nor my mailservers are hacked, take a look at the mailheaders:
Return-Path: <stasIsaev@serversniff.net>
Received: (qmail 25562 invoked by uid 0); 22 Jun 2007 12:17:06 +0300
Received: from 220.125.204.181 by post (envelope-from <stasIsaev@serversniff.net>, uid 92) with qmail-scanner-2.01
(clamdscan: 0.90/2659.
Clear:RC:0(220.125.204.181):.
Processed in 0.239443 secs); 22 Jun 2007 09:17:06 -0000
Received: from unknown (HELO ?220.125.204.181?) (220.125.204.181)
by post.ziniur.lt with SMTP; 22 Jun 2007 12:17:04 +0300
Received: from [220.125.204.181] (183.178.25.193)
by stasIsaev@serversniff.net with SMTP;
for <gvitkauskasd@ziniur.lt>; Fri, 22 Jun 2007 19:17:22 +0100
MIME-Version: 1.0
None of these sender-IPs belong to serversniff.net's infrastructure. Seems that it's time to drop the catch-all-adress for serversniff.net.
tom
Monday, June 18, 2007
For the records
For the records: Our update-lag with domainnames is at 443,017 days, and it's increasing. It will continue to increase for some time, for 450 days back was a time where we did bulk-updates: inserting many many new hosts from big lists at once, without to much handling of domains and ips at all. There wasn't too much data or trigger-overhead, and the database was not yet public. I hope to catch up the update-lag to 400 days in about a month and be around 200 days by the end of the 2007. I don't really believe than we can get smaller update-cycles with our current network-bandwidth. But you always have the option to filter outdated records from beeing displayed, regardless if you use serversniff.net, tomdns.net or serversniffs api.
tom
tom
Sunday, June 17, 2007
Serversniff deLux
If you ever wondered what serversniff looks like:
Its located in the attic upstairs from the garage, where it's hot in the summer and cold in the winter. Its made of a cheaposystem with an Athlon 3Something with one Gig RAM and an old Perc2-sc-scsi-controller ripped from a Dell-server and a Proliant-HDD-array from ebay.

I suffered occasional blackouts when lightning stroke, doing damage to the database - so i ordered a brandnew UPS, the small thingy standing right, coming straight from china. We're prepared now.
tom
Its located in the attic upstairs from the garage, where it's hot in the summer and cold in the winter. Its made of a cheaposystem with an Athlon 3Something with one Gig RAM and an old Perc2-sc-scsi-controller ripped from a Dell-server and a Proliant-HDD-array from ebay.
I suffered occasional blackouts when lightning stroke, doing damage to the database - so i ordered a brandnew UPS, the small thingy standing right, coming straight from china. We're prepared now.
tom
Saturday, June 02, 2007
crazy ideas
around 500 days ago i had a crazy idea: mapping the net in a database. all domains, all hostnames, all relations of ns- and mx-servers.
i knew a few sites who should have this data but would not really let you look it up - whois.sc and netcraft.com were amongst them. that was all i knew. oh and yes, i knew mysql, i worked with sqlite and microsofts sql-server for years.
i expected this to be an adventure. a textadventure, fun.
and hence, it was fun. the database crashed, servers got blocked, i had errors in my harvesting scripts and i was overwhelmed when i got around 100 million domains with even more hosts at once.
and now i'm sitting on a bunch of data that gets older. my time is limited, my hardware-ressources are as well. data is getting old. i started updating the hostname/IP-entries these days. I should have done this earlyier, i know - but i didn't. time is limited - remember?
a bit frustrating: the "update-lag", the time between the last update of a hostentry is currently exactly at 427,167 days. most frustrating: it's still increasing.
hmpf.
tom
i knew a few sites who should have this data but would not really let you look it up - whois.sc and netcraft.com were amongst them. that was all i knew. oh and yes, i knew mysql, i worked with sqlite and microsofts sql-server for years.
i expected this to be an adventure. a textadventure, fun.
and hence, it was fun. the database crashed, servers got blocked, i had errors in my harvesting scripts and i was overwhelmed when i got around 100 million domains with even more hosts at once.
and now i'm sitting on a bunch of data that gets older. my time is limited, my hardware-ressources are as well. data is getting old. i started updating the hostname/IP-entries these days. I should have done this earlyier, i know - but i didn't. time is limited - remember?
a bit frustrating: the "update-lag", the time between the last update of a hostentry is currently exactly at 427,167 days. most frustrating: it's still increasing.
hmpf.tom
Subscribe to:
Posts (Atom)