841
Security / Re: In reality, computer forensics are often worthless for evidence
« on: July 06, 2013, 07:46 am »
Seriously there have been exploits in the past for turning somebodies computer on the Gnutella network into a remote proxy, tricking it into downloading whatever you want it to, from whatever server you want it to, and storing it in its shared files ( which actually allows you to turn Gnutella into an anonymity network for yourself as well ) . I wonder how many people arrested for sharing CP on this network even realize that such an attack is possible. In the majority of cases I highly HIGHLY doubt that the forensic analysts (who are sometimes trained for only a few months prior to being certified) going over the seized computers realize that such an attack is possible. They are not trained to look for things like this, they are trained to run P2P spidering software and basic software forensic tools that do all of the real work for them. Guess what, anybody who went to prison for having CP obtained from the Gnutella network, or who went to prison for having CP while running one of the many vulnerable Gnutella clients, did NOT get a fair trial if they did not know about this attack!
www.ics.forth.gr/~elathan/papers/gdos.acns06.pdf
This attack can be used to get arbitrary people to download CP and put it in their shared files, and if it is carried out correctly Mr. 6 month forensic certificate is not going to be able to tell that this is what happened. Forensic analysts often are not going to be able to tell if somebody fell victim to this attack or if they actually downloaded and shared CP. And yet there are hundreds of thousands of people in prison right now for having downloaded CP off of gnutella and shared it. And probably NONE of them know about this!
www.ics.forth.gr/~elathan/papers/gdos.acns06.pdf
Quote
3.2
Exploiting the HTTP protocol
A large number of HTTP requests that result in an HTTP 404 response code
may not be difficult to handle for a Web Server. The attack can be more efficient
if we can force the Gnutella peers to perform an actual download from the Web
Server. The download may not even be relevant to their search criteria Server.
This can be achieved by embedding a specifically constructed file name in the
QueryHit packet. For example, consider that a Query with search criteria “foo
bar” is received. The file name:
Quote
We have demonstrated how unstructured P2P systems can be misused for
launching DoS attacks against third parties. We have developed an attack that
exploits a number of weaknesses of unstructured P2P systems and manages to
instruct innocent Gnutella peers to generate a significant amount of traffic to a
victim host. The victim can be another Gnutella peer, but also a host outside
the Gnutella system, such as a Web Server.
Although the basic attack relies primarily on the ability to spoof QueryHit
responses, we also took advantage of the HTTP protocol used by Gnutella peers
for data transfers. This allowed us to construct malicious QueryHits that result in
downloads of arbitrary files from a target Web Server. An interesting observation
is that the use of HTTP in this case allowed the attack to “leak” to other systems
as well.
Finally, we have developed SEALING, an algorithm which aims at keeping
a local “Safe List” on each peer, containing IP addresses and port numbers of
hosts that have been characterized as non-Gnutella participants. Our algorithm
assumes that any connection from Gnutella participants to non-Gnutella partic-
ipants is a possible DoS attack.
Quote
Indeed, with modest effort we have managed to
develop techniques, which, if adopted by bogus peers, can result in DoS attacks
to third parties by redirecting a large number of peers to a single target host. In a
nutshell, whenever they receive a query, these bogus peers respond by saying that
the victim computer has a file that matches the query. As a result, a large number
of peers may try to download files from the unsuspected victim, increasing its
load significantly. Furthermore, we have developed mechanisms which trick this
large number of peers to actually download files from the unsuspected victim.
To make matters worse, in our methods, the victim does not even need to be
part of the P2P network but could also be an ordinary Web Server. Therefore,
it is possible for a significant number of peers attempt downloading files from a
Web Server, increasing its load and performing the equivalent of a DoS attack.
This attack can be used to get arbitrary people to download CP and put it in their shared files, and if it is carried out correctly Mr. 6 month forensic certificate is not going to be able to tell that this is what happened. Forensic analysts often are not going to be able to tell if somebody fell victim to this attack or if they actually downloaded and shared CP. And yet there are hundreds of thousands of people in prison right now for having downloaded CP off of gnutella and shared it. And probably NONE of them know about this!