Parasitic Computing

b0r0din writes: "CNN has this article about a way to force computers to solve complex computational problem using the checksum algorithm used by the TCP/IP protocol. For more technical details, see their website." You probably thought learning TCP/IP was useless. No! You can use it to make an extremely inefficient computer...

Comments

Legality?

[silicon_synapse]

Such online piracy does not violate the security of hapless servers, using only areas specifically earmarked for public access, according to the researchers.

But it could slow the machines down by engaging them in mindless conversation while they unwittingly work for their remote master, Barabasi said.


Isn't this theft of resources? The researchers are literally stealing bandwidth and clock cycles. Maybe it's just me, but this seems very ethically wrong. I wonder if an IDS or firewall can be configured to protect against such leeching. Any lawyers or firewall experts in the house?

Re:Legality?

[Mr. Sketch]

I wonder if an IDS or firewall can be configured to protect against such leeching

IANAFE (I am not a firewall expert), but the only way I could think of would be to always ignore the checksum so they always get a connection and thus it would screw up their results. Otherwise it's indistinguishable from normal traffic, it's just bad traffic. Maybe the firewall could start dropping packets after X number of bad checksum packets?

This looks possible, but why?

[phoenix_orb]

This will make an EXTREMLY slow computer, and if anyone out there knows anything about routing (which I am sure you do :) The time it would take to recieve and compile all of the data would take longer and require more bandwith than would be viable on the economy of scale.

Example --- need to send 4,000,000 packets out and recieve the TCP packet back.

To do this with any speed, and also to not lose a fair majority of packets, you have to have a huge backhaul.. (T-3, OC3 or larger) TCP will not continue sending packet so you will loses them. Cost for large backhaul. $4800 month, (as by what my company chages..)

4,800 x 12 $57,600

So for one year of a huge pipe to the net you will be paying 57,600 (through my provider)

This still will not fix latent packets that never get back to the user, or any other problems.. (such as someone on your network running bearshare and eating all of your bandwidth)

Now lets look at the amount of money used for that large amount of bandwidth.

$57,600 for the amount we could have spent on that line in one year we can build a beowulf cluster with 30 nodes (and that is being very liberal on the cost of the nodes)

Now, looking at the article that I read, it seems as if the computing style using TCP/IP is very very ineffiecint.

Personally, for the amount needed to make this work, on the scale of actually getting any real work done, I would much rather build a Athlon Beowulf cluster.

This looks like in reality this could only be implimented in the real world as a new type of DOS attack.

Re:Interesting concept

[orangesquid]

Anyway, other than the TCP checksum, are there any other protocols out there that do something more computationally intense to the data before returning it?

An interesting idea is the hijacking of authorization sections of secure protocols, dispatching authentication requests based on a public/private key pair you are trying to hack, to thousands of servers and the one that returns a successful result must have been given the correct key pair.

Of course, I'm fairly certain most widespread secure protocols can't be used like this, but one or two of the less common ones might have a loophole... but then again, if they're less common, resources would be scarce, and you're better off trying to crack things on your own.

Possible application

[jmv]

Hey, I could turn CodeRed into a SETI@Home client!

Why not use ICMP echo instead?

[adadun]

ICMP echo packets (ping packets) also includes a checksum. By using the ICMP checksum instead of the TCP checksum, almost every computer connected to the Internet could be used for computation, not only web servers.

It is both legal and ethical - no questions here..

[hillct]

Although this is a really neat implementation, I'd have ot say the authors of the paper on 'Parasitic Computing' were over-reaching when they said:

Parasitic computing raises important questions about the ownership of the resources connected to the Internet and challenges current computing paradigms.

Granted it's a neat implementation, and using communication protocols to evaluate mathematics is vary creative, but what questions does it really raise about resource ownership? Vary simply, No. They're using publicly agailable services for a new and creative (not illegal) purpose. It would only become both illegal and unethical is their use had the effect of a denial-of-service attack. The authors of the paper were trying to inflate the importance of their work by tying it into the many and varied discussions of intelectual (property) and physical resources on the net. It's a shame that they took this approach because their research stands alone without such pandering which serves only to diminish the appeal of their work.

--CTH

Let the MPAA servers DeCSS for you

[Quikah]

You could write a DeCSS client to find a decryption key by sending the computations out to the MPAA's servers. :)

I Have DONE THIS! I Did it years ago(steal cycles)

[Anonymous Coward]

I Have DONE THIS! I Did it years ago (steal cpu cycles remotely for local computational tasks in a distributed network manner without having account priveledges on any target systems)

Many unitversities in the 1980s used the MERIT network and many still do.

A feture of MERIT allows logging onto any other system from another system and during a login process a free command line feature allows use of the CALC calculator line command.

This exotic command would only work for a while before they severed the line after about two minutes, unless you finally logged in validly so they could charge you the 9600 baud access fees.

The calculator command was great. It allowed a truly dumb terminal to do simple math functions. Other 1980's terminals such as Liberty Freedom Ones and other terminals have built in desk calculators modes.

You can use the calculator function to do multiplication and other operations without owning a system account. It even worked during modem connections and tou could tie up several connections by "hopping" during a login.

I created tools to use the math functions of the MERIT network to perform computations FOR FREE.

Merit is a private, non-profit corporation, governed by thirteen of Michigan's four-year publicly supported universities. In addition to the thirteen members there are 230 affiliates with a combined total of 425 dedicated network attachments from 398 separate locations. Merit affiliates include: 85 Colleges and universities,25 Community colleges ,117 K-12 schools or school districts , 22 Local, state, and federal government agencies ,16 Healthcare organizations , 111 Libraries , 21 Other non-profit organizations ,28 Businesses . Most were Amdahl mainframes (IBM clones).

Stealing free cpu cycles of innocent target machines as a parasite to perform complex computational tasks of a larger state machine, using network protocols is fun, especially if distributed across multiple systems and limitless.

I proudly did it first in the early 1980s.

(I have a life though and achieved many other more useful things by the way)

F.E.

Ideas, please!

[megaduck]

Most of the posts here have been of two schools:

1. It's impractical.
2. It's unethical.

Both valid points, but I think that it's foolish to dismiss this out of hand. First of all, it's a pretty slick hack. Very inventive, if nothing else. Secondly, it brings up some very interesting questions. Can this ever be made practical? What would it take? Would it be ethical to make it work? Can this be used to augment a DOS attack, or something similar? If so, how do we defend against it?

Maybe I'm talking out of my ass here. I don't know TCP/IP very well. However, I know that others of you out there really know your stuff. I'd like to hear from you.

Contradiction

[D. Mann]

"We are not worried about copycats taking our program," Barabasi said.

But variations could be engineered to make online piracy much more efficient, he cautioned.

If it will make piracy more efficient, I'm pretty sure the pirates would be very interested in finding out more about it.

Hell, in my experience, most pirates would use a modem that belched huge clouds of carbon monoxide and was powered by grinding up kittens in a big hopper if it got them an extra 10k/s on their downloads.

Why not enable this on purpose?

[GrEp]

Why not make this a feature? Write an extremely simple virtual machine that would perform calculations as asked. Way smaller than java. Simple enough that you could write a proof that it couldn't try to play outside its sandbox.

You could give it a small chunk of memory to use, run it at a VERY low priority, and use SSH like transmission where the packets are automaticaly compressed and only a list of certain IPs would be accepted. All you would have to do is download the IPs of the distributed projects you wanted to work on and the virtual machine would accept packets from them. No specific clients to download for each project, and you would get distributed computing easily on all your machines.

Any projects like this? It would be great to have an always on and client secure distributed computing platform.

If you do it, do it right.

[Apuleius]

For full effect, use avian transport for the
TCP/IP packets. And write an interface to this
so that you can use it for SetiAtHome.