SSL Requests sent to port 80 (request for help/input)

Published: 2012-09-06
Last Updated: 2012-09-06 23:37:12 UTC
by Johannes Ullrich (Version: 1)
3 comment(s)

A while ago, a reader submitted some odd looking web log entries like the following:

 

default 10.5.0.48 - - [06/Sep/2012:23:11:36 +0000] "\x16\x03" 200 15 "-" "-"
default 10.5.0.48 - - [06/Sep/2012:23:12:26 +0000] "\x16\x03" 200 15 "-" "-"

 

After some experimenting, we figured out that these are SSL connection attempts that are directed at a non-SSL server. These log entries are common if your web server is misconfigured, and the SSL module is not enabled on port 443. But in this case, the log entries showed up on a web server listening on port 80.

To "force" an https request on port 80, you have to add the port explicitly to the URL. For example the log entries above, I created using the command:

wget https://webserver:80/index.html

The bytes "\x16\x03" are the first two payload bytes transmitted for the connection (a \x00 is the third byte, which terminates the string as far as the server is concerned). The server may actually still respond with a default error page. The server I used was configured to respond with a "200" code for any request (to make URL brute forcing harder).

The Wireshark analysis of course doesn't make much sense here:

However, luckily we can use Wireshark's "decode as" feature to make more sense of the packet. If we ask Wireshark to decode this traffic as SSL, we do get a perfectly fine Client Hello packet:

The "0x16" byte indicates that this is a "Handshake" and the "0x03" tells us that we are dealing with SSL 3.0. So now we can do a bit fingerprinting on these requests.

Here is a sample from today's ISC log:

\t\xe2)\x18\x12\xbc\xcc\x04U\xbf\xddj\xc4\xf9q\x163\xa0\x90
\xc4]\xd6\x1cg\x90\xc1\xf2\xe9\x9a\x1e\xba\v\xca2N\x92\x1a\xd0
\xb2\xf28i\xe5{A\x16`\xc2\x01\xa1\x84\xd4_\xfe%\x93\x92\xf8\xb1
\xb7\x85\x15\x05\xdc\xae\xde\x9d\xbb'\x05\x8e\x11\x17\xb9\xdf\xee|%\xd19\xf3\x9b\xeb
\xb9\xa4\x03{\xea\x88\xf4\x88\x87\xfb\x17\xc5\x07\x9c\xc5{\xaa?{\xc7]v\xcf
\x04c\x98\xbf\x87+in
\x83\x91w\xe2\x13\x85\xae,qs\xdb\xbe\xd78\xa4\xed\xbf\
\xd2\xa2 *\xcaUV\xd7\x0e\xab\xaa\x91A\x13\xf7E\xaf\x01\xc1\x9e\xbf\xd3
\x99\xd2\xad\x1b5\xcc\x85\xef\xaa\r:9\xdc>p\xdf\xfb\xb8\xb6\xd1Pj4\x04\xb1\
\x1f\xe4\xbet\xec\x0c\xcc>\xf3\
Avg\xdc\x94\xd5\xd9\bO\x18y+\xcd\xb0
\xd1\xaf\x855\xeb\xb4\x19/3\xa8\xab\x15ZNZU9>=\x0e\x87\xb8\xa0\xe2\x12
\x16\x03
\x16\x03

(I shortened some lines a bit to avoid page width problems)

Based on our analysis above, on the last two appear to be SSL requests. The others, appear to be "something else". Can you identify them? Telnet doesn't cause any additional characters to show in the log (I suspected telnet's terminal negotiation, but was not able to trigger the characters ). Have you seen similar entries in your logs? Haven't seen anything with SSH either. The SSH client first waits for the "banner" from the server before sending anything (I may have to wait longer).

------
Johannes B. Ullrich, Ph.D.
SANS Technology Institute
Twitter

3 comment(s)

Comments

What we really need is a "magic" file for network traffic, just like what is used for the unix "file" command.
Somebody should start an open source project to create that
> What we really need is a "magic" file for network traffic

How about:-

http://www.ntop.org/products/ndpi/
It appears that someone has configured a mail exchange to your port 80 using SSL/TLS. This can happen by default if your main MX host does not reply in time and your @ for the domain is now your web server as the default behavior for SMTP is to try all MX first then try the root of the domain. Some even try port 80 thanks to Microsoft. Just a first glance guess, but it is most likely that. It could also be a generated test from a SSL or TLS test site. Best, Al

Diary Archives