The Human Buffer
Overflow - Code Execution in the HumanOS
security professionals we all love the idea of writing a good buffer
overflow. To define, a buffer overflow
is when extra data overwrites memory that may contain other data,
including variables, instructions and program flow control. This can result in erratic behavior
which can cause the program to terminate or give errors... or in security
terms cause a breach resulting in code execution on the remote server.
how can this be possible in the human brain? Can we get the human mind to
"buffer overflow" and execute code we want it to?
establish a simplistic baseline for this discussion. Try to read the COLOR of the word not
what the word spells. Do it as
fast as possible, not stopping to think.
It is not terrible, but it will illustrate how easy it is to
inject a thought... if you succeed do it faster and faster if you can.
is this so hard? It is the way the
human mind is wired. Our brain
sees the color but it reacts to the word being spelled first. Therefore the thought in our minds is the
WORD not the color. This exercise
shows it is possible to have "code" execute in the human brain
that might be the opposite of what the person is thinking or seeing.
has been proven that we speak 150 words per minute but we think at
500-600 words per minute. This
means that most people we talk to can jump around our conversations in
their heads. So overflowing the
brain through fast speech seems almost impossible.
must also understand how people make decisions in life. Most decisions we make will be done
subconsciously. We make decisions
like how to drive to work, to get coffee, to brush our teeth, what
clothes to wear without really thinking about it.
Have you ever driven all the way to work and when
you get there, you can't remember what billboards you passed, what route
you took or that traffic accident on the news? You were in a state of mind that where
your subconscious took over and did what you always do without consciously
thinking about every turn.
Most decisions we make are like this. Some scientist even believe we make
decisions up to 7 seconds in our subconscious before we make them in the
real world. When we finally do
make a decision consciously we do it from more than just what we hear...
we get our sight, our feelings, yes, our emotions involved in the
Understanding how humans work and think can be the
quickest way to creating our buffer overflow.
Fuzzing the HumanOS....
Just like fuzzing a program, where we through different
lengths and pieces of data at the program till it crashes, we have to
understand how the human mind reacts to certain types of data. There are
certain laws in the human mind that seem to be inherent and we all follow
them. If you approach a building
with two sets of doors and you hold the first set open for a complete
stranger what do you think they will do next? Either hold the next set for you or
make sure that set stays open till you get inside. Or if you are in a line of merging
traffic and you let a complete stranger merge in front of you... most
likely if you needed to merge later on he would let you in without even
The Law of Expectations - The law of expectations basically
states that a person will usually comply with an expectation. Decisions are usually made based on
what that person feels the requestor expects them to do. This is one way
we can start sending our malicious "data" to the brain
Presupposition can be described best by giving an
"My next door neighbor Ralph always drives a
green ford escort."
In this sentence we presuppose:
I know my neighbor
His name is Ralph
He has a license
He drives a green car
To use presupposition effectively you ask a
question using words, body language and facial expression that indicates
what you are asking is already accepted.
The basic gist of this is to bypass the "firewall" (the
conscious mind") and gain access directly to the root of the system
The quickest way to inject your own code is through embedded
The rules of embedded commands
There are some basic principles to embedded
commands to make them work:
Usually they are short - 3 to 4 words
Slight emphasis is needed to make them effective
Hiding them in normal sentences is the most effective use
Our facial and body language must support the commands
Embedded commands are popular in marketing with
things like "Buy Now", "Act Now" and "Follow
me" are common ones. Like any
good shellcode we need some padding to help the commands find their
way. We can utilize phrases that
will help the target become padded to our code. Phrases like "When
you...", "How do you
feel when you...", "A person can..." , "As
you..." - all of these statements create an emotion or a thought
that allows you to inject code into the subconscious.
Let's give an example... if you were on a sales
call and wanted to use an embedded command you might say something like:
you purchase a product like this from
someone like me what features
are most important to you?"
The bold letters are where emphasis is placed but
the sentence causes the person to use their memory to think about a time
they bought this product and what is really important, as that thought is
occurring you are injecting your code "When you purchase from
What's important is to not mess up our tones... if
we are OVER emphasizing the words then we will sound odd and scare the
person off instead of embed commands.
Just like a normal buffer overflow the information must match the
command we are trying to overflow.
Putting it all together
As you probably have already imagined this is a vast
field with a large room for messing up.
Practice will be needed to be very successful at it. Although we do not promote using this
information for seduction there are some decent videos that show how
embedded commands can work.
We do not want to paint the picture that this
becomes some Jedi mind trick. That
is not the case. Just because you
tell the person "You will purchase from me" does not mean they always will. So why use these commands?
It creates a platform to make social engineering
easier. It is also a good lesson
to companies you work with to educate them what to look for and how to
spot someone who may be trying to use this information against them.
So let's write out the equation more simply:
Start off with phrases, body language and speech that is
assumptive. Presume the things you
ask for are already as good as accomplished.
Next pad the human mind with some statements that make it easier to
embed code, at the same time embedding code. In essence this is the recipe for the
human buffer overflow. Use it
sparingly, but practice A LOT before you attempt it. Try it at work or home. Use simple exercises like trying to see
if you can get a fellow employee to serve you coffee. "Tom, I see you are heading to the
kitchen, will you get me a cup of coffee with 2 creams please?"
Escalate it to larger tasks to see how far you can get. Try to use this to get commitment from
people. Eventually use this to see how much information you can get, how
many commands you can inject.
Let us know how it works for
By Chris loganWHD Hadnagy
What's in your
When the Internet bubble was
still rolling, vendors used to come by a lot. Almost daily. It got old
pretty quick, and if you were sitting through the multitude of sale
presentations, you would have to come up with something to keep your mind
active. My friends and I made a game out of the situation, and would try
to see how much we could learn about the sales person by studying their
desktop. It was amazing what we could put together by just paying
attention to the taskbar and desktop icons.
You can get a very complete idea
of what sort of person you are dealing with by the simple things like
their desktop environment. For instance, what brand of computer are they
using and do they have the default crapware still installed? What sort of
antivirus are they running? Instant message client? Mail client? VPN?
Perhaps they are running a remote administration client.
Let's look at an example: http://webdev.ccac.edu/talkin/desktop.bmp
Here we see they are using a
palm pilot, have a old version of Netscape and acrobat icons installed on
their desktop, leading us to believe they have this software installed.
The Outlook icon appears to be the mail client on the system, and based
off the icon we can see that it appears to be Office 2000. Based off of
just that limited amount of information we know quite a bit we could use
to craft an attack.
To look at something a little
bit more complex, let's look at an archive of shared desktops: http://www.flickr.com/groups/lifehacker-desktop-showandtell/pool/
This will going to be a bit
more challenging, as we have to expect that the people sharing their
desktops on here have highly customized them. There are a number of
screen shots here however, so feel free to try it on your own.
The picture at: http://www.flickr.com/photos/islanddog/3992811883/sizes/l/in/[email protected]/
is a good example of another one for us to analyze.
can see based on the UI that it appears to be Windows 7. This is
primarily based on the taskbar changes, but we can't say with 100%
certainty as there are some applications that will change Vista to look
like Windows 7. We can assume that the user is going to be running a lot
of "new" software as they are a early adaptor. That is driven
home by the fact we see Office 2010 in the recent applications list.
Microsoft Live messenger is in use, and we are able to see the username
in place for this system. Firefox is installed, but not running, IE
appears to be the default browser. A Pirates game is installed, and
Filezilla is an FTP client installed.
just this information, it is simple to start building a profile of the
user as a tech savvy early adaptor. We have the users IM screen name, and
the fact they play a game. Perhaps a social engineering attack could be
conducted with this information, delivered through IM.
brought this back to mind for me was preparing for a presentation I had
to give. I found myself looking at my desktop and thinking "If there
is time that my slides are not on screen, what does my desktop tell that
audience about me?". There are a couple of ways a social engineer
can work this situation.
one is try to prevent information leakage. Perhaps set up a new profile
for the purposes of presentations that is used only for that purpose.
Then the audience might be able to gain some information about your
system, but very little about you directly.
two would be to try to manipulate the audiences view of you as the
presenter. What sort of image do you want to portray, and what should
your desktop look like to support that image? How should you set your
desktop image, a solid color or a image? What applications should you
have on your desktop? For instance, if the goal is to appear to be very
professional and trustworthy, having a P2P client as a frequently used
application may not be a good idea. On the other hand, a corporate
desktop pattern might be great support.
sort of casual information leakage happens all the time. Desktops images
are just an example, and I am sure you can come up with many more. A
social engineers, we have to be aware of where we might find information
leakage and how we can put it to use. Plus, taking advantage of our
information leakage to support various pretexts we are using is a asset
that is too good to not put to use.
Written by Jim