I shouldn’t need to stress the importance of sanitising user input on web forms. I also shouldn’t need to stress this importance of government websites being secure.
I also shouldn’t need to stress the insecurity of client-side code.
However, it seems Cadw (“the historic environment service of the Welsh Assembly Government”) seems to be stuck a bit too far in the past before people started exploiting websites for fun or profit, as I recently discovered from this tweet:
The worst possible way to guard against SQL injections: http://www.cadw.wales.gov.uk/ [see source and then cry]
— Tim Dobson (@tdobson) March 28, 2011
- Adding functionality
Both of these points obviously need exploring further.
The client has the option to execute it or not, or even to modify the code first then execute it (with a little know-how).
Several things immediately spring to mind:
- SQL keywords and syntax in a “bad list”:
"select", "drop", ";", "--", "insert", "delete", "update", "char(", "`", "varchar"
- Weird stuff, possibly passwords or other language constructs?
"/", ":", "?", "|", "declare", "convert", "@@", "2D2D", "4040", "00400040", "[", "]"
- “xp_” – perhaps a computer name prefix for systems running Windows XP?
Addressing the points in reverse order, a quick bit of poking (just a standard HTTP HEAD request!) the web server reveals:
Trying 18.104.22.168... Connected to www.cadw.wales.gov.uk. Escape character is '^]'. HEAD / HTTP/1.1 Host: www.cadw.wales.gov.uk HTTP/1.1 200 OK Date: Wed, 30 Mar 2011 04:29:53 GMT Server: Microsoft-IIS/6.0 X-Powered-By: ASP.NET Content-Length: 20796 Content-Type: text/html; Charset=ISO8859-1 Set-Cookie: ASPSESSIONID......GDEM; path=/ Cache-control: private
(I’ve removed the actual cookie set :P)
Ooh look! We’re running IIS 6.0 as the web server. This gives is two likely suspects for the operating system of the server: Windows Server 2003 (aka WinXP server edition), or Windows XP Professional x64 Edition. Basically, XP.
With only talking to their web server, I’ve now got a likely prefix on machine names – chances are the names are just numbered after that, and given their network is running Windows servers, it’s likely to be on a Windows domain. That simple knowledge gives me the hostname of a large number of workstations: xp_1.cadw.wales.gov.uk (or maybe xp_01.cadw.wales.gov.uk or xp_001.cadw.wales.gov.uk, or perhaps even xp_01.wales.gov.uk etc). It would be trivial to find out which of these naming schemes existed – probably by just pinging their DNS server.
At this point, this information is getting scary. I’d like to remind my readers that everything I have done so far, I have documented here. I have done nothing else. I’d also like to remind folks that this is a government computer system, and any vulnerabilities I find I am not going to touch, as I don’t have permission to do so. Information I have found so far is either public information that they may or may not have inadvertently published (such as POTENTIAL machine names), or information that would be retrieved by software such as web browsers every time you loaded the site. Getting that information manually by simulating a (poor and slow) browser just happens to be easier than messing around inside my browser (chrome) config at the moment (for firefox users, the extension Firebug will nicely show this information for you). If you choose to use the information I have published here, then you do so at your own risk. My aim in this is to point out bad security practice in the hope that others will heed the warnings and not make the same mistakes.
The weird stuff which makes up my second point could be anything, a bit of googling might tell you why they’re dangerous, or explicitly prevented.
Lastly, the first point. Let’s take a look at the main items from the SQL-specific part of their “naughty words”:
"select", "drop", ";", "--", "insert", "delete", "update", "`"
(I’m going to quickly point out they convert the words to lowercase to check them against the list.)
So, we can’t retrieve or modify the data. We can’t delete data from the table, but truncate table isn’t restricted. We can’t use comments. We CAN use quotes, but not the table-style backtick quotes. We can’t drop tables or columns, but we could add new tables and columns if we wanted.
The fact that only the backtick (`) is in the list could be an indication of a style of quoting, which we could make use of.
Oops, did we just turn off all protection against SQL injection attacks on your database for ourselves, with a simple checkbox in the browser settings? How inconvenient of me!
Usability for tech noobs
One last point just to round off the whole thing, one on usability.
Let’s say I want to search their site (using their tiny search box) for “how do I select a place to visit?”, my search query gets cut off at “…vis”. Assuming the user is smart enough to realise the computer doesn’t like long searches, they might rephrase to “how do I select a place?”.
Would you understand it if you were a tech noob? I doubt it. After all, what’s wrong with “select”?
YOU’RE DOING IT WRONG.
CADW: If you ever see this post, fire your web developer, and take your site offline until it can be fixed by someone who’s actually competent.
Please, please, PLEASE let this be a lesson to other people how sanitising user input is a Good Thing.
I only hope that the Government in Westminster hasn’t made the same mistake, or this could be very costly… it appears someone has filled in the census form in a rather interesting way…