I'm pretty sure the first letter was G, but maybe it as a 2.
April 18, 2013 12:23 PM   Subscribe

How do police departments use/search partial license plate information in their investigations? Or incomplete information about cars generally?

The events in Boston have gotten me wondering how the police track down leads, and track down leads with respect to cars in particular.

For instance, say someone sees a car speeding away from the scene of a crime. It was a blue SUV, and all they could catch was the first three (of six) characters on the license plate. The witness may or may not be confident in their interviews what the characters were, but assume say they emphatically "remember" 012. But say it really was CIZ (or any number of alternative combinations). And say the witness wasn't sure what state's plates it was, but the lettering wasn't red (but might have been blue, black, or green).

Is that enough for the police to find the car's registration? Do the police automatically run plates based on every permutation of potential character (C/O/0/Q,9/6/G,1LI, etc), in every possible state with the right color plates? Is this done by some computer somewhere, or by brute force detective work?

And how much investigative weight is put on license plates, given how easy it would seem to make passable fakes (I remember making plastic ones at a kid's birthday party a million years ago), or steal them from other vehicles?

Alternatively, it seems like the police would get a lot of incomplete info about cars in general (it was an SUV, I think, er, maybe a station wagon, and it was blue, black, dark green, but maybe brown, I'm not sure)--do they actually run searches on that basis, or do they just keep it in their back pockets as corroborating information as other leads unfold?
posted by Admiral Haddock to Law & Government (5 answers total) 3 users marked this as a favorite
Data point: in the 2010 Times Square car bomb attempt, the bomber had used a license plate he stole from a car at a junkyard in CT. So the authorities definitely do follow up on leads based on license plates.
posted by reptile at 12:45 PM on April 18, 2013

When I worked at the phone company, we had a tool where you could put in whatever numbers you wanted in the phone number (First three, last 4, middle 6) and it would give you a list of available telephone numbers with that combination. It was used for generating telephone numbers with particular number combos, or that spelled things.

I came up with 888-Tax-1040.

I also came up with 888-Buy-Sell

That was a million years ago when dinosaurs roamed the earth.
posted by Ruthless Bunny at 1:32 PM on April 18, 2013 [1 favorite]

I suspect that people qualified to answer your question can't without endangering their job. Giving out real specific information on how things get solved can help bad guys up their game and get away with more.

When I processed insurance claims, part of my job was to look for red flags and decide if something looked fishy enough to refer to the fraud department and let them decide if it genuinely merited investigation. One phenomenon that goes on with insurance fraud is they send in something small and see if that gets through. They do that to learn the procedures for submitting claims and test the system. So folks who want to do bad things go looking for data on things like that. People trying to stop them don't really want to publish a playbook on the Internet for how to more effectively get away with it.

I have a certificate in geographic information systems, which is in part training in a kind of specialized database. Processing insurance claims involved accessing various databases. Because of my educational background, I was better than a lot of people in the department at digging through the databases and finding my way around their quirks and limitations. This was time consuming and I couldn't always spare the time. But the point I want to make is that human skill, knowledge and interpretation can make up for database limitations, assuming someone has the time to brute force it.

In other words, even if the database does not have an algorithm for looking for a zero and other things that might be partial matches for zero (like C), a human can run multiple searches, manually typing in all the different permutations and partial matches they can think of. I have done things like that to find a matching policy or whatever (like if a claim was submitted with partial information and couldn't be readily matched to the correct policy).

I don't know what the DMV databases look like or what their search parameters look like. Whatever they look like, they exist and someone devotedly trying to catch the bad guy can go the extra mile, do the research on (say) which states have green or blue or black lettering, run the individual permutations that the system may not do automatically, etc. To a large degree, the difference between a human figuring out the permutations and a system coded to do it automatically is a matter of time. A person doing it manually takes more time. (But there can be downsides to a system being coded to do certain things automatically. Every single time they upgraded our system, they introduced all new bugs. I sometimes wished they would just leave it the hell alone.)
posted by Michele in California at 3:41 PM on April 18, 2013

Searching a database using fuzzy matching is straight forward, but the database has to support it.

As far as I know, there's no federal database of all US license plates. License plate databases are maintained by the separate state MVAs/DMVs, and accessed by law enforcement through various systems. If you can believe realpolice.net, not all MVA databases support(ed) partial matching. There are other databases besides those maintained by DMVs, though. A license plate lookup during a traffic stop would involve criminal information database- whether the plates are stolen, whether there are outstanding warrants, et.c. Criminal information database systems can be state, or federal (e.g. NCIC or IBIS). Interstate/international queries may go through NLETS.

Another source of license plate information is that acquired by license plate readers, which OCR passing license plates and look them up. ANPR/LPRs are now widely in use, courtesy of DHS grants, and some automatically provide fuzzy matching. LPRs can record passing license plates, allowing real time location, or they can flag vehicles based on a hot sheet. (pdf)
Because of the speed of the reader systems, the volumes of plates being read and the fact that an alarm must occur within seconds to be useful, LPR systems use a large list of target plates stored locally in a “hot list” rather than relying on real-time communications with State or Federal data sources.
The witness may or may not be confident in their interviews what the characters were, but assume say they emphatically "remember" 012. But say it really was CIZ (or any number of alternative combinations).

Some jurisdictions try to be somewhat proactive about this:
Several states do not regularly use certain letters — most commonly the letters I, O, and/or Q — in their plates, except on vanity plates, so as not to confuse observers with the numbers one and zero.
TL;DR: it's going to vary from state to state, system to system, how well-trained the police officer is, and how motivated they are about tracking the plate down. Many states are still working on integrating their different databases.

Further reading:
For interstate traffic records, you'd check the national driver register, per the Driver License Compact/Driver License Agreement.
NIEM is a national data model used to facilitate information exchange between agencies.
posted by zamboni at 4:48 PM on April 18, 2013

Zamboni gives some good resources.

But the best answer is the simplest one- they find all the plates that match the known pattern(s) and then start to eliminate the ones that can't possibly be right. Once they have a couple of possible hits, they use standard investigative techniques to try and match the known facts with the potential suspects.

Sometimes that search means sitting an officer at a terminal entering in all possible combinations, and other times there are tools to do the job. Even if the database is relatively stupid, a macro-like tool isn't that hard to build.

The difficulty is in getting connectivity. (And this is what the DHS is trying to do on their IT end) Some states have no way of entering into their system except from a physical location. So law enforcement has to buy a leased line and a terminal and do their lookups that way. Or there will be a law enforcement liaison in the DMV agency that answers calls from non-local law enforcement agencies and generates reports for them. As systems get more and more modern, it becomes easier and easier. But connecting people with the databases they need, especially when it is not a regular part of their work, is a huge hassle in governmental IT.
posted by gjc at 3:56 AM on April 19, 2013

« Older HR Interviews and Unorthodox Employment History   |   Cooking Classes in DC for my friends Newer »
This thread is closed to new comments.