Ants Vs. Worms — Computer Security Mimics Nature

An anonymous reader writes with this excerpt from Help Net Security: "In the never-ending battle to protect computer networks from intruders, security experts are deploying a new defense modeled after one of nature's hardiest creatures — the ant. Unlike traditional security devices, which are static, these 'digital ants' wander through computer networks looking for threats ... When a digital ant detects a threat, it doesn't take long for an army of ants to converge at that location, drawing the attention of human operators who step in to investigate. 'Our idea is to deploy 3,000 different types of digital ants, each looking for evidence of a threat,' [says Wake Forest Professor of Computer Science Errin Fulp.] 'As they move about the network, they leave digital trails modeled after the scent trails ants in nature use to guide other ants. Each time a digital ant identifies some evidence, it is programmed to leave behind a stronger scent. Stronger scent trails attract more ants, producing the swarm that marks a potential computer infection.'"

Comments

ridiculous references

[sopssa]

What's with the ridiculous reference to ants? If they had said this in a technical way, I might actually even understand what they mean. Now it's basically "ants travel inside your network". The article doesn't tell a lot more.

Obviously nothing is "traveling" inside your lan cable. So do they mean they have every machine in promiscuous lan that tries to seek what is traveling there? What kind of "scent" does it leave when it detects some threat and how are the other computers interact with that?

Stop doing some stupid nature references just for the hell of it, give technical details.

Re:ridiculous references

[Chris Burke]

And ant colony algorithms by themselves are just an obfuscated way of defining ad-hoc probabilistic algorithms.

It's not "obfuscated", they are explicitly in that class of algorithms.

In other words it's completely heuristic, there is no actual theory that justifies defining the algorithms in that particular way.

Yeah and there's no theory a priori justifying simulated annealing or genetic algorithms work in their particular way. But they work. Random heuristics work (and there is theory explaining why, in general, they do). Different heuristics have different properties that are beneficial in some circumstances. That's why there's more than one. That's why Monte Carlo simulations weren't the final word on random heuristics.

And what they do has precious little to do with actual ants.

As much as genetic algorithms have to do with biological evolution or simulated annealing has to do with cooling metal. As in... next to nothing in a literal sense, but quite a bit in an inspirational or metaphorical sense. Probabilistically following previous paths through the solution space, with those paths 'evaporating' over time unless reinforced, is a pretty good analogy for what ants really do, and a good hint as to the algorithms advantages -- it does a good job of finding and tracking changes in the solution space in dynamic situations.

Basically, "Oh it's just a heuristic and not literally like the inspiration its named after" is the worst way to dismiss an algorithm ever.

Though, on the other hand, why this is a good idea for network security, I don't know. Why would you want a bunch of agents to "swarm" a location where problems are found, rather than just, say, deleting any instances of virus/worms found, and closing any security holes found (or notifying the sysadmin so they can), is beyond me.

Obvious questions.

[palegray.net]

The second question depends heavily on the answer to the first.

• Who gets to decide what qualifies as malware or a "threat?"
• Why should user agents trust this assessment?

Taking the analogy further...

[AdamInParadise]

In nature, an ant can get infected by many kinds of fungus, and when they return to the colony or meet another ant, the fungus can spread to another host.

Similarly, deploying this kind of "digital agents systems" opens another path of transmission for viruses and worms.

It's nice to see that some people are still active in this research area, but does anyone knows of a product that actually use such a principle for real?

Bound to fail

[Tinctorius]

Taking the obvious problems with this approach aside (using viral programs to identify viral infections), it should be easy to distract the flock of "ants" by one or more decoy infection(s), and then start the 'real' infection on the "other side" of the network. The "ants" have built a highway of warning signs towards the decoy(s), so the probability of ants traversing to the 'really' infected machines is lowered.

It's always fun to apply theories from one field of CS (namely optimization) to another (security), but if you give it a short thought, you know this can't be a good idea. It wouldn't be science if they didn't test that hypothesis, but I certainly hope they're not that stupid to test it in production systems.

How long before malicious ants will appear?

[misnohmer]

Having anything "crawl" through your network seems like a huge security risk to me. Any security solutions will have be aware of those crawlers and allow them to crawl from computer to computer. What's to stop viruses to simply impersonate such crawling ant - free pass to every computer on the network!

Another problem may be as they all "converge" on threats. What is they bug down the target machine, or the network? If my browser cookie looks "yummy" to the "ant" (no pun intended - browser cookie may be classified as a threat), next thing I know my network interface is crawling with these "ants"! My administrator cannot log in because of all the ants plugging my bandwidth!

like in nature there's always a hack: Phorid flies

[garompeta]

The genus "pseudacteon" of the Phorid flies zombifyies ants laying eggs in the ants thorax. The larvae moves to the head of the ant and it feeds itself until it is big enough to come out, decapitating the ant.

So yeah, I think I know how this story of swarming ants are going to turn out.