Everybody is related to everybody at six degrees ... so they say. So starting with Osama bin Laden, at some shallow depth of link analysis, you are connected. This notion begs the question: When starting with a bad guy, what kinds of links between entities (people and organizations) are worthy of attention?
Got links?
Not every link is as useful as the next. For example, when analyzing newspaper stories it is not useful to simply consider close proximity names a link. Al Qaeda and Condoleezza Rice appear to be connected over 2,000,000 times according to Google. Heck … Google also says I’m connected to Mohamed Atta over 100 times. And since you are reading my blog, my apologies, now you too are connected to Mohamed Atta at just two degrees.
Other more solid connections can be useful in some settings, not others. If you have ever seen a compiled public records report, the type purchased by private investigators, you would quickly note that at least half the named people on your report are people you do not know. All of the neighbors on your street. The person that owned your car before you. The people who lived in your house or apartment before you (college dorm rooms are a classic example). You likely know few, if any, of these people. And while very useful from an investigatory perspective (e.g., a parental abduction or murder investigation), such connections are less meaningful in the context of predicting the next terrorist.
Noting these observations in the early 1990’s, I chose to focus on a more narrow connection … a connection I would refer to as “relationships” – specifically, the likelihood that two people know each other in a close, personal sense. For example, when inventing software to protect casinos from “tightly held conspiracies to do evil” (see inverse post here), it became self evident that not only must one start with a bad guy but also pursue relationships in a very narrow manner. Employees who handle cash who are roommates with gaming felons present some risk. Employees would be expected to disclose such. Is this a telltale sign of a criminal intent or a crime? Not in the least! Is this something worth a little more attention than my mom? Well when it comes to casinos, and their expected levels of due diligence, the answer is yes.
What kind of data proves useful in expressing a close personal relationship? Well this generally involves either shared resources (homes, cars, phones) or personal communications (e.g., calls, emails, care packages, money wires). There are a few others, but I will have to let your mind wander as I would hate to tip off any evil doers.
Even when starting with a bad guy, and following only close, personal relationships, the usefulness of the trail still degrades very quickly. That is unless the trail leads to another previously known bad guy … then of course, those in between are certainly a bit more interesting. Link analysis brings with it many interesting national security policy questions. To name a couple: How deep should a government agency be permitted to see? And if there is another evil doer at the other end, does that provide probable cause to see the entities in-between? And it better be much less than six degrees, huh? Otherwise, all paths lead to you, me and Mr. Arbitrary.
I think one of the things people need to understand is that you can't just blindly apply link analysis. You need to analyze the content of the links to make sure if there is something useful. Therefore, there needs to be some type of semantic analysis behind the links and filtering before going in depth further with the link analysis.
Posted by: GadgetMan | May 16, 2006 at 10:03 AM
Greetings expensive Jeff Jonas, work with Technology of Computer science and I live in Brazil, also I admire the work that you develop in its country, would like to know if it has possibility of work integration here representing its services in Brazil. Since already I am thankful for its atenão, peace and happinesses.
Jayron Soares
Analyst of Support Technician
Posted by: Jayron Soares | June 21, 2006 at 07:49 AM
I design the help file for a Social Network Analysis program and found you while doing a web search. For the most part the data they give me to work with is difficult to understand since I don't know the whole situation.
Instead I created a set of data from the tv show, "Stargate SG-1". Since I watched the show for ten years the dataset is much closer to me and the reports I run make much more sense. I can also detect when there might be problems in the actual data.
It is the most fascinating subject I've ever worked on.
Posted by: dave columbus | September 25, 2007 at 05:01 AM