A Modest Proposal for Homeland Defense
By David Holtzman

“… the threat of terrorism is an inescapable reality of life in the 21st century. It is a permanent condition to which America and the entire world must adjust…the need for homeland security, therefore, is not tied to any specific terrorist threat.”
The Presidential Report on Securing the Homeland

 

What went wrong with our intelligence system? Why didn’t it give adequate warning in advance of September 11?

While those questions have not yet been completely answered, the current administration is currently busy at work planning a new information system – one capable of anticipating terrorist acts using artificial intelligence, or “data profiling” technologies.  Predictive systems are what intelligence analysts call "Indications and Warning" and require orders of magnitudes more information then normal monitoring or tactical intelligence systems.  They also require substantial guesswork.  Computer scientists call these "heuristics", analysts call them "WAG"s.

So how would you go about building this type of system?

I've outlined seven technical steps that will probably have to be done.  All of these have been accomplished on a smaller scale, so it shouldn't be a problem. 

1.  Start by connecting existing government and commercial databases

The first step would be to connect most of the large government databases that contain information on domestic activity—including customs, immigration, law enforcement, military records, IRS files and eventually state and local tax rolls, political contribution lists, educational and voting records. In the short term, they would build software that translates queries between the various databases since government information systems are currently the digital version of the Tower of Babel. The permanent solution would be to create rigid requirements requiring all agencies and contractors to converge on a common set of standards for data storage and access.

Contractors would eventually be hired to write translation gateways into many commercial databases so that searches against the government database could also be seamlessly integrated. Some of these commercial databases would be straightforward, such as credit reports, phone and other utility bills, and transportation/reservation systems from airlines, rental car companies and hotels. Many might be more subjective in nature and involve human appraisals such as profiled direct marketing lists, school guidance counselor records and comments made by utility or government workers who have access to individual residences.

2.  Match them to commercial information like credit reports using the SSN

Initially the translation systems would be “data-matched” against the government records, but it’s inevitable that the government would eventually mandate that all commercial databases have a social security number field. This would likely result in legislation making it a crime for consumers to give a false social security number to companies and might even require these companies to deny service to curmudgeons that refused to provide their SSN.

3.  Add tens of millions of cameras and other sensors

This system would eventually access tens of millions of real-time sensors for up-to-the-minute threat assessment. This process is already underway by several different agencies. Some of these sensors include visual cameras at various public places such as storefronts, street corners, highways, toll roads and airports. Some of these cameras are already utilizing experimental face recognition software. Other sensors would include identification devices at checkpoints to public buildings and eventually all transportation terminals. 

4.  Require a national id card and tie it into a biometric database

These devices will require some sort of universal identification card coupled with biometric information. These biometrics could include fingerprints, retinal scans, face measurements, blood type and DNA. The military is already collecting DNA information to facilitate body identification. Of course this will require a national ID card and even more importantly a universal database of biometric information otherwise it will be useless. The easiest way to build up this database is to collect the information from school children. An alternative method would be to link the biometric collection to draft registration for citizens and visa issuance for resident aliens.

5.  Track phone calls and email and generate diagrams of social groupings using traffic analysis 

An important element of a predictive system will be information on social interactions and “networks” of individuals who communicate as a group. Intelligence analysts refer to this process as “Traffic Analysis”. Expansion of the Carnivore program to encompass most internet-based communications coupled with records of phone transactions should provide enough information. Unlike phone conversations, the cost of examining all communication is not cost or legally infeasible. 

Naturally, detailed analysis will also require the content of the conversations. Since the system will have to reconstruct activity after the fact, this implies that all communication from all Internet users will have to be stored. This includes chat type communications, emails, web content and file content because a process called steganography can be used to insert messages into say, the background of a picture.

6.  Build technology that will "guess" what people are thinking and what they might do

Since terrorism is ideologically based, anyone might potentially be a terrorist, and therefore everyone’s actions will be have to be under constant scrutiny under this proposed system.  The biggest problem with large-scale information systems is figuring out what’s important amongst all of the data that’s being stored. Since this is a threat assessment system, the process of assessment is generally accomplished by “heuristic” processing or rules-based analyses similar to those used by scoring systems that take credit reports and apply a set of rules to them to determine creditworthiness. A rule might decrease a credit score because the consumer hasn’t lived at the same address for more than two years for example. A terrorist profiling system would have much more sophisticated and insightful rules because they would be crafted by psychologists and would have much more data to work with. These rules would look for either ideological leanings as demonstrated by choice of reading material, organization memberships and friends or psychological disturbances as evidenced by behavioral changes such as switching buying habits at the grocery store. 

The attractive part of heuristic systems is that new rules can be tried out at very little incremental cost because the interpretation takes place on one government computer in the software and doesn’t require new sensors to be deployed. An unlimited number of these systems can be used in parallel against the same data, which gives the government the ability to experiment with psychographic profiling without causing any public notice since the cost will be well below the radar screen of the budget appropriation process. Researchers will be free to experiment with many types of correlations of individual behavior such as dietary habits, travel behavior and social grouping to determine the best way to predict people who may in the future become threats.

7.  Give everyone a secret "THREAT score" or "LOYALTY rating" using technology similar to that used by credit card companies 

Since millions of government workers need access to these threat profiles and most of them will not be trained in the nuances of interpreting psychological information, the most useful way to display the results of these profiles will be “Threat Scores” which are similar to credit scores. This way any government employee with access to the system could look up a person’s Threat Score based on their Social Security Number, driver’s license or Immigration Visa number and treat the person accordingly.

People will get used to the cameras, but the threat profiling will cause them to make lifestyle adjustments

Most people are flexible to changing notions of privacy and as they adjust to life in this virtual fishbowl, they will quickly realize that any deviation from “culturally normal” behavior as defined by the system may raise their Threat Score. We’re already conditioned for this kind of behavior in many ways such as knowing that reporting an accident claim raises your insurance premium or that if we leave a job too soon it can affect our credit rating. We’ve become accustomed to the idea that our credit report can affect our ability to get a job, rent an apartment or buy a car. The Threat Score would serve the same function in all of our interactions with government employees. Most Americans will be very motivated to maintain a low Threat Score and if that means giving up a little individuality, they’re willing to do it.

As this Predictive Information Security System develops, people will quickly find out what kind of behavior will draw attention and what’s safe. They’ll naturally tend to avoid certain books and take extra special care to find out the background and opinions of their friends, colleagues and employees. Since the concept of what kind of behavior is threatening will always be changing, but the data trail is permanent, many citizens may find that the prudent thing to do is simply to avoid calling any attention to themselves for the duration of their life.

If someone unfortunately gets a high Threat Score, perhaps because of something that one of their friends or family said, it would be useful to provide a mechanism for decreasing the score through some socially useful action such as providing information on one of their neighbors.

• Respond on the Message Board
• Back to Previous Page