UCB, USC To Build (And Hack) A Model Internet

darksoulz writes "Associated Press stories from TheKansasCityChannel.com and TheChamplainChannel.com have an interesting report today. It appears that the U.S. Government has given a $5.5 million grant to the University of California, Berkley and the University of Southern California so that they can build a model of the internet, so they can hack it. They are trying to find better defenses against hacking, without breaking the real Internet. The first phase is scheduled to be completed by February."

Comments

Good

[delirium28]

This is an intersting approach, but if they keep this a "closed system" (i.e. a hidden internet), then how much real "hacking" can they expect? Are they going to use honeypot data? Will they expose this system externally for real "hackers" to play with? Just my $0.02 worth.

Re:Good

[Anonymous Coward]

I think the real point here is to make a testbed where they can unleash worms and then try different techniques to try and trace, contain, stop, and prevent them. Not to actually attract hackers. They will be the hackers, do the hacking, etc. Also to play with DDOS attacks and whatnot, without saturating any existing networks. I think the operative term is hacking a model 'internet' not a machine on the internet.

Are they doing the FULL internet here?

[paroneayea]

I mean, will sixty percent of it be model porn?

I'll do it for 1 million

[the_skywise]

Why do you need 5 million for that? How many computers can it possibly take? 50? 100? Let's say 100. That's $100,000 (and that's generous these days) Let's say $200,000 to lease building space and power for 2 years (also generous) and let's pay 3 professors part time, plus 10 students work study wages (Figure $50,000 per professor and $20,000 per student...$350,000)

What do we got?
$100,000 parts
$200,000 space
$350,000 labor
--------
$650,000

What's the other ~5 million going for?

Oh wait... they must need Windows licenses and full copies of Outlook to properly test the hacks...

Re:I'll do it for 1 million

[dr_dank]

What's the other ~5 million going for?

Mountain Dew doesn't come cheaply, my friend....

Re:I'll do it for 1 million

[Anonymous Coward]

Sorry but you're pricing things according to how much they look like they'd cost to you. Not how much they'd cost to the people who are actually buying them.

If you get paid $50,000 a year by your employer (before taxes), how much do you think you cost to your employer? $50,000? Ha. Try $100k. They play various kinds of employment taxes, insurance, maybe a bit to a personnel management company... a good rule of thumb is that you cost 2x your salary to an employer. (Plus, you're taxed again on your end: a 50k salary means your take-home pay is 35k or whatever.)

Also, at a university, you have what's called "overhead". If a research group gets a $1million grant, how much of that grant money do you actually see? Well, at my university, the answer is something like 45%. The university takes a huge rakeoff of 55% to pay for buildings, infrastructure (e.g. networking, plumbing), deans, blah blah... that's where the university's income comes from.

So, I'm sorry to tell you, that if Fry's sells a computer for $1500, you need a much bigger grant to buy it. Try $6k.

In ten years...

[sandbenders]

In ten years everyone will wonder why USC and Berkley produced all of the decade's best crackers. This project will result in three things:

1. Good dissertations for CS PhDs.
2. More secure software, which will rarely be implemented and even more rarely be implemented well.
3. A whole bunch of research assistants who think they are l33t h4x0rs. And some of them will be right.

Re:Is it dead already?

[nate1138]

My old roommate used to say that only two good things ever came from Berkeley:

• BSD
• LSD

Of course he went to Cal-Tech (Which I've always heard was a trade school for surfers ;-)

Questions remain

[bwaynef]

I think this is a straight-forward approach but there are problems that need to be addressed.

• How will they implement the latencies and delays that commonly befall the internet.
• Where will the millions of unique traffic-data originate.
• How will they mimic the different styles of hacking.

It would be interesting to see how they implement this network.

• Are they going to be two seperate networks.
• Will they be one network connected over an existing internet link?
• If they connect over an existing internet link how will they deal with that connection as its no longer a simulation

I don't think that we have a requirement to see any of the information that I've questioned above but this information could lead /. to be more informed on this situation.

Just pretend don't hate me...

[Dareth]

... but can I be the RIAA and sue all the users of your model internet???

This is in the name of science!

I won't be real bad, just demand they hand over all their old video cards when they get new ones.

I'm still running my old Voodoo 3 3500, yeah baby it still ROCKS!

The Real Reason

[fireboy1919]

Tech support companies the world over are tired of people calling up and asking, "Is your internet better than the internets of other companies?" They made a plea to the US government to do something about it, who, out of embarassment for the American people's stupidity, promptly made up a new reason to make another, lower quality internet.

From now on, rather than spending several hours trying to explain the concept of the internet to people who have trouble walking and chewing gum at the same time, tech supporters will be able to simple say, "Yes."

Just offer prizes to hackers

[G4from128k]

I'm not sure how they plan to "model" the internet, but I would argue that the internet is its own best model. Anything else will lack some exloits present in the "real" net while have other exploits absent in the real net (bugs in the model's software).

I would take the $5.5 million and divide it up into $5000 prizes that are payable to any hacker that demonstrates and documents a hack on the real net. The profs and grad students could ajudicate the prize giving. They would find at least 1100 exploits this way (fewer if they have to pay those pesky grad students or usurious university accounting department overhead rates).

If letting hackers profit from hacking the actual internet is too scary/illegal, then the university could create a small publically exposed network running a variety of apps, OS, etc.

Similar Project at Iowa State University

[logrey]

Iowa State has a similar project funded with a $500,000 grant from the U.S. Department of Justice.
Iowa State Computer Security Lab

The nature of government grants

[mystery_bowler]

For the last few years I've been developing software systems for law enforcement, so occassionally I pick up interesting bits of information about how government funding works. If you didn't hear about it - and not many people did - the Dept. of Homeland Security made a sort of "open call" (via the Dept. Of Justice, if I remember correctly) about a year or so ago. It was - more or less - an open invitation for vendors to propose innovative ideas to the DHS about fighting terrorism within the United States. The really interesting thing about the open call was that it was specifically worded to encourage "innovative" and "new" approaches. I joked at the time that I actually felt good about the open call...it seemed like the guys at the DHS were acknowledging that they didn't have a clue what to do and where looking for expert help on making things radically better.

I'd be interested to find out if the "model internet" was a proposed idea. In terms of government funding, $5 million isn't all that much, so I wouldn't be surprised to see if this was an idea pitched by people at UCB and USC during the open call. I'd heard that big names asking for reasonably small amounts of money were getting through pretty easily.

I tried to convince my company to pitch a variant of our crime analysis/trendspotting tools. Include a reference per recorded crime that indicates political or religious bias as the motive of the crime. Get a concentration of those - even if they are "lesser" crimes like vandalism or simple assualt - and you've got "smoke". And where there's "smoke"...

Re:Tin foil hat or not?

[ShinmaWa]

I'm absolutely positive that part of the agreement was that Berkeley could publish their findings once the project was over. If absolute secrecy was important, then it wouldn't be done at Berkeley (and it wouldn't be reported on Slashdot). Its that simple.

Besides, in order to "shore up the US", there would need to be a lot of cooperation among the multitudes of private enterprises that actually run and maintain the network. Most of the big players in this arena are large, multinational companies that would be inclined to shore up ALL their networks, not just the ones in the US.

here's the real scoop

[t_parker16]

all this speculating on what's involved, but the project is described in pretty good detail over at the ISI web site. (and so, its apparently not USC specifically but the usc information sciences institute):

http://www.isi.edu/stories/70.html

excerpt:

"The DETER testbed will consist of approximately 1,000 computers with multiple network interface cards, located off the actual Internet. Three permanent hardware clusters, or nodes, at UC Berkeley and at ISI's Southern California and Virginia facilities, will serve as the core of the system.

"This isolated mini-Internet will serve as a shared laboratory where researchers from government, industry and academia can test existing and new security technology, using a wide variety of attack techniques."

Re:This will fail

[orangesquid]

What makes you think they would pick a good operating system on purpose? Rather, they could put up many different systems known to be hackable, write worms or scripts designed to hack into these machines, and try to create technologies to capture/contain and lessen/slow infections and security breaches.

I don't think the point is to re-create OpenBSD. The goal is probably more of a cross between network monitoring, intrusion detecion systems, and automatic network reconfiguration.

The Internet can already route around problems because of redundancy. Sophisiticated routers can control and shape traffic. But, as of yet, there's no widespread technology to protect entire networks from security problems. We will never create perfect systems... so we must create countermeasures so that when our systems fail, they fail in the smallest and least dangerous ways possible. It's like fault isolation.

a model?

[hyrdra]

The problem with their "model" is that something as complex as the current Internet as it exists today can't really be modeled, at least not very well. It's a huge chaotic system thats constantly changing and growing, so when you try to model it your model is going to be out of date before you can do anything useful with it. I really don't think $5 million can buy even a small representation of what the Internet is today. Think of the OSI layers and all the different software, hardware, protocols, methods, systems and manufactures in place at each layer. Each of those has its own set of vulnerabilities, holes, etc. and keep in mind there are many different versions of each of those running at the same time across different networks.

This complexity is precisely what makes tracking and solving problems with today's Internet so hard.

I am curious as to what they expect to study and find from this model. Today's problems with the Internet and networking in general are largely social, economic, or political. Figuring out some neat new protocol isn't going to make backbone provider X update their entire network. Worms and the such are also the bain of a social problem. As long as we have smart programmers with free time, there will be worms and exploits of the system no matter what procedures are in place or how smart the network is; The fix for said worms are timely patches and updates, however most users won't and don't do this, hence the epidemics.

This might make some great academic research and a neat new toy for the University but I fail to see how it can find applications in the real world where the problems are much harder than the technical ones this project (presumably) hopes to solve.