The Death of AV Defense in Depth? Revisiting AV Software
It has been quite some time since I updated this blog, I will try to update the blog in the next weeks, with a few details what I was up to during the last months.
Let's start with the more important stuff, I got into AV Research again =) The output of which will hit the public in the next months, be warned there will be a flood of advisories :D
Together with Sergio Alvarez I gave a talk @ Hack.lu 2007. This year we explained what the heck is up with Anti-Virus software. We revisited the way AV solutions are implemented in current Company networks, and AV Engines themselves. Defense in Depth is being misinterpreted and incorrectly implemented with disatrous effects. (They believe they do DiD when in reality they do not, this is an important fact to keep in mind.)
Rough Break-down of the Talk :
DiD as implemented for Anti Virus Software is broken, companies put one AV engine after the other believing it to be DiD. The worst security incident in such an architecture is being incorrectly defined as "A virus passes the gateway unrecognized" , in reality the worst possible failure is that the underlying Operation System is compromised through the AV Engine, you have to mitigate this.
AV Software is broken behond recognition, they parse enormous amounts of Data in unmanaged programming languanges and such are naturaly prone to errors. This was clear from the start, but the shear amount of bugs is someting else.The reality shows they all are.
AV Software runs directly on critical (with high privileged rights) infrastructure, AV Software runs everywhere
E-mail changes what is at stake: What happens if I sent an exploit targeting AV software as an attachment in an E-mail ? (You can automatically compromise Corporate Mail Servers/Clients/Gateways, from the outside as your email travels through your firewalls untouched
This is kind of old news and present it here for the archives, a German TV Channel "SAT 1" did a documentary with me on Bluetooth security. I am not happy about the outcome, they had around 5 hours of footage, 5 minutes were used, and only the lame stuff shown. The real stuff (MiTM,cracking etc) has not even been indirectly refered to.
Sight, anyways here it is :
For those that understand German here is the archived footage :
My call for an OSS Bluetooth sniffer during the
last 23C3 in Berlin has not been unanswered, first there was Max Moser
("Bluetooth - Getting raw access") that uncovered how you can
modify a consumer USB stick by flashing it with a commercial BTSniffer
firmware (there was at least one vendor that included the firmware with
every trial download) and get RAW access to it.
The question that was left
was how to send commands to it, get it into sniffing mode, synching it
to the other devices. Exactly this is what Andrea Bittau and Dominic Spill
found out during his work on a Paper entitles "BlueSniff:
Eve meets Alice and Bluetooth", he further implemented it in C code.
The paper will be shortly be published and presented at this years' USENIX.
In other words a Bluetooth Hacker dream has partially come true, a cheap and (partialy) open way to sniff and capture packets, including the Pariring-handshake which may than be cracked.
Andrea is currently working on cracking open the very last thing that holds him from crafting low level Bluetooth packets, the XAP2 processor, he dissassembled the firmware to find out how exactly it works, for that he wrote his own dissassembler. After this he/we may write our own firmware and basicaly do whatever we like, for example a full blown fuzzer or full blown attack device.
Other very interesting findings will be uncovered, more on this later :)
I repeat: "Islamic Terror is the number one threat to Europe" is the chanson
of lots of politicians, now check this :
My question is this, how comes a politician can make such
claims without being held accountable, when there is clear evidence that
he is simply pushing an agenda and doesn't care about reality ? Not too
mention these other "Terrorist" groups always existed, always were there.
One thing is clear to me, if we don't get off our lame asses and start
to do something we soon will be no better than the US.
BTCrack 1.1 is
ready! I named it BTCrack Heisec release, because I released it during
the Security Conference of Heisec
BTcrack is a pairing handshake cracker against Bluetooth
1.0 - 2.0
for more information please resort to the Paper by Shaked and Wool
and the website listed at the end of this E-mail.
In cooperation with PicoComputing (http://www.picocomputing.com/)
we added FPGA support to BTCrack 1.1 and increased the
Software speed by 15% reaching 200.00 keys per second on a stock P4-Dual
Version 1.1 :
[+] Added Priority Control
[*] Fixed splash bug
[+] Added FPGA Support
[+] Speed increase (15%)
A friend and colleague of mine, namely Alexios Fakos
has published a Book under the title of Sichere
Web Anwendungen, unfortunately
it is german only. If you'd like to know how to code hardened Applications
I heartly recommend this Book.
Kevin Finistere and
myself gave a Bluetooth Presentation at the 23C3 congress
in Berlin on the 29.12 at 14:00 local time. We released a bit of 0day and
a bit of protocol bugs and tidbits. See for yourself :) Thanks
to everybody that made this possible also thanks to the CCC for
organising this event, while I couldn't really participate as a spectator
at least I can judge about the behind the scene work. I was impressed.
The organisation was good and poeple very friendly and helpfull.
Please for those writing about this lecture, they
key findings are
not the tools. You have been so spoon fed by the last Bluetooth
talks that you think what is important are tools :(
There now are underlying protocol security issues not only pure
implementation issues :
What is important to understand is that non-discoverable
mode no longer represents a protection. Naturaly before attacking a
devicey ou have to know it's there, previosuly you could try to bruteforce
the address (bd_addr) in order to find it, if it was in non-discoverable
state. This however is a unreliable and takes a very long time.
During this talk i released information on how to PASSIVELY
discover 90% of the address and bruteforce the 8-bit remaining. Which
is reliable and fast.
THIS means the the only protection Bluetooth
has to protect you from connecting to my device is GONE.
Also the paradigm shift from toys to workstations is als
We can eavedrop on your Laptops Microphone, we can compromise
it and take control over it.
The random number generators and the encryption affected by them is
New Re-pairing attack, making the pairing attack INTERESTING.
Your drivers lack control, no update function and are flawed beyhond
Live demo on how to take over a PC and get a remote shell over Bluetooth
from the Lecture :
Pin and Link key recovery is practicaly possible (code release and
If you use Bluetooth beyboards or mice, your PC has a HID server, these
may be attached to inject commands (!) as if you were typing on the keyboard
The random numbers used for encryption and so forth may be very weak
for your device
The Pin is not that usefull the Link
key is !
Swap over to Bluetooth 2.1 (as soon
as possible) and use "Secure
Regard the quality of the encryption Bluetooth offers (E0) as a PRIVACY
feature NOT a security feature. (Compare it to WEP)
New re-pairing attack : Connect to the master pretending to be from
the piconet, use a fake linkkey, master will think (oops lost the pairing)
and will re-initiate the pairing given an attacker the choice to capture
the exchange and crack it.
Don't trust encryption taking place, sometimes the
devices negotiate Security Mode 2, and you don't know your data is
actually transferred in clear text (after being authenticated) and you
can't actually check as you don't have a Bluetooth Sniffer.
The Bluetooth PIN is actually a Bluetooth Passkey, it supports
characters not only digits (this has security implications)
Things to do once you have the link key:
Passively decrypt the traffic
Connect to the slaves
pretending to be the master and have full access (no pin required)
Connect to the master pretending to be one of the slaves have full
access (no pin required)
Plant the link key on a BT capable machine and have a remote encrypted
stealth channel to that machine
Update your Drivers !
Widcomm, Toshiba, Bluesoil, ALL vulnerable
Don't rely on Windows update for that, your BT stack may be from
a third party vendor (very likely)
Listening on the Microphone and recording is also possible on
PCs (not only cars)
General Recommendations :
Delete your existing pairings as soon as you don't need them
Pair in "secure places" SIG
As soon as your device asks for
a PIN again, don't enter it you
might be snooped on (see previously
mentioned pairing attack)
Don't trust Bluetooth 1.0 - 1.2
(can't tell for 2.0-2.1 yet)
Mitigate and Monitor.
using Bluetooth for Industrial
Regenerate a new key every 5 minutes, use 16 chars.
PLEASE implement the GUI to use the possibility for bluetooth
to use characters (UTF8) NOT ONLY DIGITS.
Please be more transparent towards your device driver version
numbers and propose an easy way to update.
BTCrack 0.9a is
going ahead nice, optimisations have been done and the final release
will be on the Nruns website as promised very soon™ :)
BTCrack 0.9a now spawns 8 threads in
order to crack the keys, and this implies that dual-core or
quad-core processors are working out very nicely :) A few assembler optimisations
are still ahead and the final release should be ready for 23c3.
The general assumption that the attack is theroreticaly
possible and that
Pins of 6 digits represent a good protection is now pratcticaly
Here are my current stats on a Dual-Core P4 2Ghz (48000
pins per second)
over! a nice security conference in Luxembourg. Had a great time, although
sometimes organisation was a bit messed ;) Reeaaaallly nice and very interesting
poeple, commercial rate was very low and finaly I saw some poeple I knew
only virtually in real life.
Well as you knew or not knew, Kevin
Finistere and myself
gave a talk about Bluetooth security. Yaaaaaaaaaaawwwwwwwwwwnnnnnn ?
I started using this tool last year ago during internal
tests, it was immediately obvious to me that this is a great tool to
have. It's name is Satori,
if you never heard about it that's not a proof the tool is no good but
rather that it's Author Eric
Collman does not really seem to care if you do (or at least doesn't
scream it from the top of every house)
I found out about Satori while reading the paper "Chatter
on the Wire" (from the same author) which goes into great
length about passive OS fingerprinting and it's potential for improvement
as done by several other tools. What is interesting is that the paper
was not only theoretical but rather practical, it's outcome was Satori,
a beautiful plug-in based Passive enumeration and Fingerprinting tool.
Satori uses Winpcap and captures packets passively at
the NDIS level, every packet flying by is being scrutinised for information
that might determine it's OS. Nothing new here you might say, well Satori
does the fingerprinting on :
DHCP, BOOTP, ICMP, TCP, CDP, EIGRP, HPSP , HSRP, HTTP, ICMP, IPX, SMB, SNMP,
STP, UPNP precisely enough to either correlate the results with nmap or to
rely on them. It makes spotting potential vulnerable systems a breeze.
It's obviously very handy for critical networks where
you are not allowed to scan or to scan only a minimum. (This does exists.)
It shows it's strength when used in internal networks,
I was able to spot machines that didn't belong in a certain critical
network immediately (as they broadcasted their Netbios presence) by only
using passive means. It's also very usefull when doing quick scans (nmap
port 80 as example) across an internal network, it gathers all packets,
makes a list of all responding machines, fingerprints them and gives
you an exportable list. Very handy.. and speedy, I was able to pump 8000
packets per second thorough without any lags or problems.
XAMPP - Multiple Privilege Escalation and
XAMPP is an easy to install Apache distribution containing MySQL,
PHP and Perl. XAMPP is really very easy to install and to use -
just download, extract and start. In the FAQ we
read : Xampp is not meant for production use but only for developers in a
development environment. However I have seen it being used in production
environments quite a lot, hence this advisory. According to the download
stats, Xampp has been downloaded 2.765.443 times between 2003 and 2006
Title : Xampp - Multiple Priviledge Escalation and Rogue
Ref : TZO-062006-Xamp
 Priviledge Escaltation to SYSTEM due to FileZilla Service
Path specification - CVSS Rating
 Priviledge Escaltation to SYSTEM due to MySQLadmin Path specification
- CVSS Rating
 Priviledge Escaltation to SYSTEM due to CGI Path specification - CVSS Rating
 Rogue Autostart due to unsecure File execution - CVSS Rating
HD Moore released a new fuzzer to the public, it's called CSS-DIE and
as it names implies it fuzzes most Browsers to death using only CSS tags. I helped
a bit on that project and HD credited my for that (Thanks!) . Thanks to
Eric Sesterhen for helping me out on the asm part.
Here are a few interesting results I obtained :
Internet Explorer 6 (latest version, all patches applied)
EIP 7C912F17 ntdll.7C912F17
MOV EBX,DWORD PTR DS:[EAX]
Opera (latest version) :
Lots of Null Pointers (Denial of Service)
EIP 67B3DF80 Opera_1.67B3DF80
67B3DF80 8B43 04 MOV EAX,DWORD PTR DS:[EBX+4]
EIP 67B3DF80 Opera_1.67B3DF80
67B3DF80 8B43 04 MOV EAX,DWORD PTR DS:[EBX+4]
EIP 67B3DF80 Opera_1.67B3DF80
Firefox has long been considered Spyware hardened and spyware safe, it never
really was. Don't get me wrong on this, it's not the fault of Firefox (although
it could be a bit better protected against this particular attack). I made a
small movie demonstrating this particular Proof of Concept.
Update: A bit of clarification what this
fuzz is all about, as
you see in the small animation, the Extension installs without
any user interaction.
That should be quite new, Firefox tries to block silent installs though random
profile directory names and various other tricks. The adbar sends any URL
you visit to a google syndication server thus monitoring your surf behaviour.
Update : The animation takes some time to
load, wait for it.
"Safe'n'Sec is complex data and user applications protection against threats
and vulnerabilities for individual PC as well as workstations in corporate networks.
The program uses proactive technology based on activity analysis in user PC. "
Multiple Insecure File execution and Autostart handling
During Startup, snsmcon.exe spawns the
GUI process named safensec.exe through the use of CreateProcess() . By doing
so it omits to set the variable'lpApplicationName' and further omits to quote
the path in the variable "lpCommandLine"....
During Autostartup, Safe'nSec omits the quotes
around the path to the executable and as such may spawn a rogue application
instead of the appropriate Starforce application.
The vendor (Starforce) did not care to respond to my report. Thus I decided
to publish this low-impact vulnerability. Update: Starforce quickly fixed the issues after the diclosure.
(see Read more)
F-Secure AV - Anti-virus Bypass and Buffer Overflow
Flaws in the way F-Secure software handles ZIP and RAR data compression archives
could allow an attacker to execute remote code on users' systems and also to
bypass F-Secure's antivirus-scanning capabilities.
I found mutliple vulnerabilities within various AV Engines, F-Secure are
the first to actually publish a real advisory, others fixed the bugs silently
or put a small notice in a change_log. I will however not
publish more details about the findings as of yet, there are too
many AV engines vulnerable to similar issues and I am going to wait
until most of them have patched the flaws until I exactly dislclose my findings. http://www.f-secure.com/security/fsc-2006-1.shtml
As employees become more mobile, sophisticated VPN solutions are required to
meet key security challenges such as securing access to corporate resources
and protecting remote desktops. To meet the VPN client needs of any organization,
Check Point offers VPN-1 SecureClient.
During Startup, the SR_Watchdog spawns the GUI process named SR_GUI.exe through
the use of CreateProcess() . By doing so it omits to set the variable'lpApplicationName'
and further omits to quote the path in
the variable "lpCommandLine"....
I decided this being not worth reporting to the vendor, this
is low impact, although it shows quite a bad coding practice.
Introduction : WehnTrust is a Host-based
Intrusion Prevention System (HIPS) that provides secure buffer
overflow exploitation countermeasures. While other Windows based intrusion
prevention systems are only capable of working with a pre-defined group of applications,
WehnTrust's technology allows it to work with virtually all software products.
Perhaps best of all, WehnTrust is currently free for home use.
Wehntrust forgets to correctly quote the
autostart key and thus may start c:\program.bat|exe|com on reboot...
Only taking into account vendor confirmed vulnerabilities.
I won't disclose what products are affected by which "vulnerability"
until the vendor has patched the software and given me clearance to go ahead.
It's amazing how differently AV vendors react to such reports,
some like Fortinet respond very quickly disclose all the versions vulnerable
and have a patch ready in merely days, others give you a single point of contact,
act very professional, give you weekly updates, yet others say they have no
time, or consider your research "doubtful" or question the "methology"
(without fixing the actual bugs).
I will include these reactions as comments and as neutral as human possible
in my future advisories.
* The number depends on cooperation of the vendor
to disclose all vulnerable versions.
please don't. I have seen tons of recommendations to filter all ICMP
traffic, I have seen poeple feeling all warm and fuzzy inside because they
blocked all those nasty ICMP Datagrams, some having severe
network congestion problems _because_ they filter every ICMP packet.
Well here is news for you: ICMP is there to help you.
Types to consider not to Filter :
- ICMP Type 3 Code 4 :
Fragmentation needed, but DF bit is set (Outgoing)
Why ? Path MTU discovery
- ICMP Type 11 : Time exeeded (Incoming)
also important when routing loops occur. Block
Block all the other ones, if you like to ping allow 8 0 out
and 0 0 in.
F-Prot/Frisk Anti Virus bypass - ZIP Version Header
Yet another AV bypass
F-prot engines failes to decompress ZIP files which have a version header
greater then 15. The consequence is that the F-prot Engine is unable to scan
the virus/malware inside and consequently flags
it as harmless. If used as an Email Gateway solution the offending Emails will
slip through. Read more
is my version of the the Bluetooth Sniper weapon, it features a medium sized
YAGI antenna combined with a 10* magnification scope and a metalised parabole
which bundles the Bluetooth signal further enhancing the range.
The interior is made from a Linksys USB dongle soldered to the yagi and
to a USB connector. Read
bug is similar to the bug reported
by Dr. Peter Bieringer, most vendors have since fixed the bug in
their ZIP unpacking functions, however a similar Bug exists in the RAR unpacking
code and probably also in other Archive unpacking functions.
Virus scanners fail to correctly scan the files due to escape sequences within
the filename and give them a clean bill, for a scanner used on an AV mail
gateway this means the file is passed on without any warning.
Companies relying on Anti Virus scanners on the email gateways
as their sole protection. These can be easily bypassed using these techniques. Read
more (Errors found, needs to be redone)
R&D section opens with
this entry : "...It is possible for Microsoft to
monitor and data mine search terms even when those are searched over Google
or other Search engines. This is possible because
Windows XP SP2 first submits the serach terms to the MSN server and then
redirects to the requested search engine.." Read
More to come soon...
OSVDB is an independent and open source database created by and for the security
community. The goal of the project is to provide accurate, detailed, current,
and unbiased technical information on security vulnerabilities. I encourage
you to join the effort and to contribute
to the Project.
posted a few pictures from my Defcon 13 visit.
Las Vegas and Defcon 13 was great, however I have been disappointed by the
level of some if not most presentations, some of them were perverted into some
sort of one-man-show not really disclosing the essence of the talks but turning
into drinking games, which may be entertaining to watch sometimes but not if
you traveled over 14 hours in a plane just to be there. Greetings Efugas by
the way. Defcon 13 Gallery
my card number was part of the "30 million visa card hack", however
unlike many other VISA offices, the repsonsible center in Luxembourg (CETREL)
immediately blocked the cards, notified the users and immediately issued new
cards. Lack of scanning device I took a picture of the letter in French and "English".