Slashdot Mirror
MIT Creates Tor Alternative That Floods Networks With Fake Data (softpedia.com)
An anonymous reader writes with word that MIT researchers "created an alternative to Tor, a network messaging system called Vuvuzela that pollutes the network with dummy data so the NSA won't know who's talking to who." Initial tests show the systems overhead adding a 44-second delay, but the network can work fine and preserve anonymity even it has more than 50% of servers compromised.
More wasted bandwidth!
This is potentially good for an obfuscated messaging service, not an encrypted internet proxy for all traffic.
Silence is a state of mime.
I wonder what % of Tor servers are compromised. I abandoned Tor when I realised it didn't mix in junk traffic like this, as traffic analysis through compromised nodes/routers is such an obvious vulnerability that it seems to render Tor worthless.
I was talking to my Google-employee brother the other day and voicing my prediction that 'virtual camouflage' would become a defense against data mining and spying, similar to as described in the article. He thought the idea was ridiculous, and even if it were to come to pass, would be defeated by statistical means. Regardless, secure p2p communication is an arms race, and the virtual environment closely resembles nature in unexpected ways.
MIT was once the number one non-profit Department of Defence contractor in the nation. Don't know how much funding they get these days but it certainly seems as though this solution is provided to you by and for the U.S. Government.
This is just like in Neal Stephenson's novel Anathem. Except when the system became fragile, the noise was mixed with the signal so most communications became worthless.
Any bozo could write random garbage and waste bandwidth. Write something that can split encrypted data at the client through multiple nodes and recombine encrypted packets at the server. And make it an IP level protocol! Idiots!
i wash my hands with a firehose
The sarcasm in Your comment is too obvious. The Innocent have the most to hide, relatively speaking.
First time I heard vuvuzelas was at Ronald Reagan's nomination at Republican convention. They were pretty effective at drowning out everything for 10 minutes or more.
44 second latency? Isn't that problematic?
I'll just leave this here: https://en.wikipedia.org/wiki/Chaffing_and_winnowing
So they're just tooting their own horn.
I eat only the real part of complex carbohydrates.
If this is designed to flood a network with junk data to conceal the relevant data, could this be interpreted as a form of a denial of service attack if it decreases network performance?
I see this as the proverbial "big stick" to push back against the conglomeration of TLAs and communication oligarchies.
"You don't want strong encryption? Then we will do this!"
Bent, folded, spindled, and mutilated.
Generating random message traffic to thwart message analysis and hide true communications is an old trick. It's really a form of steganography, just not a very efficient one. By participating in one of these networks, you draw suspicion.
People who really want to communicate clandestinely probably just use public forums and image sharing sites as digital dead drops for steganographically hidden messages. There are many steganograhic systems for a medium of your choice, many of them even auditable and open source.
Just get Netflix, Amazon, Hulu, etc... to stand up Tor exit nodes. Chum the pipeline with things like Gigli and The Last Airbender and let the NSA filter through all that. Maybe they'll just kill themselves - I know I would.
It must have been something you assimilated. . . .
Why not just encode data in normal bittorent transfers?
Normal clients would just ignore unknow data.
This idea of using padding to stymie correlation and confirmation attacks has been discussed a few times in the Tor mailing lists. It's even in the Tor Project's FAQ: https://www.torproject.org/docs/faq#SendPadding.
The main reasons it hasn't been attempted are:
- Requires a lot of bandwidth;
- Some users and relay operators wouldn't be able to operate relays and use Tor because of all the bandwidth consumption;
- Doesn't impede all kinds of correlation and confirmation attack attempts, and may open up more ways of attack; so all the overhead may be pointless some significant portion of the time;
- Latency is absurd, would make Tor nearly unusable;
- Would require significant re-engineering of the protocol at this point;
> pollutes the network with dummy data
probably not so different from internet as we know it, isn't it ?!?
Right, I'm not very knowledgeable about encryption, so please bear with me. Here's some thoughts.
Some US agents recently pressured a library to shut down their TOR node, and they did.
That raises some questions: firstly - was this a one-off random event where they just felt like it this one time, or was it part of a systematic plan?
Let's presume there was a systematic plan behind it. That should mean that they systematically pressure the shutdown of TOR nodes. So: How many times have they succeeded, and why have we not heard anything about it earlier? Correlate that with the miniscule cost for a government of running even 1000 TOR nodes, and it's quite possible that far more than 50% is compromised. Doesn't TOR presume that someone isn't controlling all the entry and exit points?
So alternate suggestion: it seems to me that the biggest barrier for running TOR is the bandwidth. We even have some people torrenting through TOR. The main parties likely to run a TOR node will be institutions and those who can smuggle a server into their employer - and these targets are those most susceptible to pressure. So how about an anonymous chat network?
Metadata is another issue - if A is kept under surveillance, and his ISP detects a connection to B, then it doesn't matter that the connection is encrypted. Once B is identified as the recipient of encrypted communications then B could be attacked through many other ways.
So how about a chat network that autonomously connects to a number of other clients, and mimics human behavior? Maybe a central database of client signatures maintained by a known privacy advocate which it downloads, and then connects randomly to each of these, transferring gibberish? It could then intermix meaningful messages to named clients with the gibberish spam, and the recipients could name which signatures it should actually save messages from. Sometimes transferring messages through random third parties as well. So A's client sometimes connects to B directly, and for periods of time randomly picks F and sends messages to F containing another message with orders to forward to B.
Who's talking to *whom*.
This has been done before -- I'm surprised at the lack of a reference. MIT CSAIL has really gone downhill since the 80s.
See: P5 : a protocol for scalable anonymous communication by Sherwood et al (2001)
This kind of thing only works if you keep it a secret. If MIT has not already been served with a national security letter you can bet that one is on the way.
See subject: Shortest path/route work - & it does work (you pick it up in discrete math usually in CS work or math majors, OR business degrees as well)... right (or wrong)?
* It came in VERY handy for me while I worked for a major frozen food distributor in the mid to late 90's in Atlanta Georgia where it was used a LOT in routing trucks efficiently as possible.
APK
P.S.=> I'd almost be willing to BET you employed that kind of math... apk
Some US agents recently pressured a library to shut down their TOR node, and they did.
That raises some questions: firstly - was this a one-off random event where they just felt like it this one time, or was it part of a systematic plan?
It wasn't. That were just some guys from the local police and the library put the TOR exit node up again.
Let's presume there was a systematic plan behind it. That should mean that they systematically pressure the shutdown of TOR nodes.
There always is and always was such pressure from law enforcement, for obvious reasons.
Correlate that with the miniscule cost for a government of running even 1000 TOR nodes, and it's quite possible that far more than 50% is compromised. Doesn't TOR presume that someone isn't controlling all the entry and exit points?
Possible but not that easy, because TOR nodes and TOR exit nodes are in many different countries. Some of them might cooperate, others certainly not. It seems like a better and easier plan to systematically compromise node endpoints themselves with trojans, viruses, or other special surveillance software.
So how about a chat network that autonomously connects to a number of other clients, and mimics human behavior?
I don't understand why you want to mimick human behavior, since the network needs to be encrypted anyway. It's better to connect all nodes in the network (or all neighbors, or random nodes, etc., depending on the network topology) by continuous low-bandwidth encrypted data streams sending and receiving at the same rate all the time. In this encrypted stream is random data interspersed with the messages.
If I'm not mistaken, such a system was or is used by embassies of one or more nation states. I think I've read about that somewhere.
There are some problems with this idea for ordinary "end consumers", though. Unless almost everyone would use this system, every node would be very suspicious and someone would invariably try to compromise the nodes with side-channel attacks. Moreover, just the location and owners of the nodes would probably be enough information to figure out who and why someone is using this system. And since PCs are totally insecure and easy to compromise, this network would probably only (politically) hard to compromise if millions of people would install and run the system. It's the classic chicken and egg thing that also prevents widespread email encryption.
But it's still a nice project, and chat is a good application for experimenting with p2p networks.
My 2 cents as AC. :-)
It's a perfect name for it, regardless if it works. Tells you exactly what it does... "WHAT??? WHAT?!?!?! FUCKING VUVUZELAS!!!" https://www.youtube.com/watch?...
See subject: & I agree - sometimes, you hit 'traffic jams' (choke points) in REALITY that the math can't 'see' beforehand... "been there/done that" too.
APK
P.S.=> What amazes me, particularly in computing when it's applied, is the work of no doubt many decades (if not a lifetime) from the "math freaks" out there (or theoretical physics too) once the applied folks (engineers) begin experimenting with applying it - "WE ALL STAND ON THE SHOULDERS OF SOME SERIOUS GIANTS" imo @ least... apk
Source code available here: https://github.com/davidlazar/...
That system seems to require a lot of random data. What is the plan to gave good enough entropy sources so that it is not broken by being predictable?
Sure, this would confuse/use up resources on an inline monitoring device. So what. The NSA collects all of the data for offline processing. It can eventually be filtered.
AC
Nullsoft's Waste (2003?) had a network saturate feature that did just that.