Thursday, September 23, 2004

Dark Networks

Vladis Krebs has a case study page examining how mapping social networks and understanding their properties can be used to take down of terrorist networks. Network analysis was used to take down Saddam Hussein. The Washington Post has some of the details.

The Army general whose forces captured Saddam Hussein said yesterday that he realized as far back as July that the key lay in figuring out the former Iraqi president's clan and family support structures in and around Hussein's home city of Tikrit.

Following a strategy similar to that pioneered by New York City police in the 1990s, who cracked down on "squeegee men" only to discover they knew about far more serious criminals, Maj. Gen. Raymond T. Odierno said his analysts and commanders spent the summer building "link diagrams," graphics showing everyone related to Hussein by blood or tribe.

While U.S. forces up to then had been preoccupied with finding "high value targets" from the Bush administration's list of the top 55 most-wanted Iraqis, Odierno said those family diagrams led his forces to lower-level, but nonetheless highly trusted, relatives and clan members harboring Hussein and helping him move around the countryside.

And the rest as they say, is history. John Robb took at look at the September 11 network and analyzed its characteristics. The Mohammed Atta network had evolved under Darwinian pressure until it reached the form best suited for its purpose: to conduct strategic attacks against the United States of America. Robb concludes that a cell of 70 persons will answer to the purpose, yet be sparse enough to allow its members to remain in relative isolation. For example, no one member of Atta's cell knew more than five others. Moreover, the average distance between any two members was more than four persons. Crucially, but not surprisingly, this disconnected network of plotters maintained coherence by relying on a support infrastructure -- probably communications posts, safe houses, couriers -- to keep themselves from unraveling.  Because security comes at a price in performance and flexibility, Robb arrives at an astounding conjecture: you can have small, operationally secure terrorist groups, but you can't have large, operationally secure cells without a state sponsor.

Distributed, dynamic terrorist networks cannot scale like hierarchical networks. The same network design that makes them resiliant against attack puts absolute limits on their size. If so, what are those limits?

A good starting point is to look at limits to group size within peaceful online communities on which we have extensive data -- terrorist networks are essentially geographically dispersed online communities. Chris Allen does a good job analyzing optimal group size with his critique of the Dunbar number.

His analysis (replete with examples) shows that there is a gradual fall-off in effectiveness at 80 members, with an absolute fall-off at 150 members. The initial fall-off occurs, according to Chris, due to an increasing amount of effort spent on "grooming" the group to maintain cohesion. The absolute fall-off at 150 members occurs when grooming fails to stem dissatisfaction and dissension, which causes the group to cleave apart into smaller subgroups (that may remain affiliated).

Al Qaeda may have been able to grow much larger than this when it ran physical training camps in Afghanistan. Physical proximity allowed al Qaeda to operate as a hierarchy along military lines, complete with middle management (or at least a mix of a hierarchy in Afghanistan and a distributed network outside of Afghanistan). Once those camps were broken apart, the factors listed above were likely to have caused the fragmentation we see today (lots of references to this in the news).

His last paragraph is crucial to understanding why the defeat of the Taliban in Afghanistan and the toppling of Saddam Hussein may have cripped global terrorism so badly. Without the infrastrastructure of a state sponsor, terrorism is limited to cells of about 100 members in size in order to maintain security. In the context of the current campaign in Iraq, the strategic importance of places like Falluja or "holy places" is that their enclave nature allows terrorists to grow out their networks to a larger and more potent size. Without those sanctuaries, they would be small, clandestine hunted bands. The argument that dismantling terrorist enclaves makes "America less safe than it should be in a dangerous world" inverts the logic. It is allowing the growth of terrorist enclaves that puts everyone at risk in an otherwise safe world.


Here's a link to a database of terrorist incidents called, MIPT Terrorism, via the Neophyte Pundit. I'll look into the site later today or this week, but it seems useful enough to put on my blogroll.