Recently the US FBi has ahad some embarrasssing failures of its high-tech computerized fingerprint matching program. the system is supposed to be able to macth fingerprints with a national database of the prints of known criminals. Supposedly the system was to deliver matches (within hours) to local police, military police, etc., and the matching was to be far better than that done by humans.
This reminded me of a story i read years ago. the german national police had invested BILLIONS in a computer system called PPP (people, places, possibles) which was supposed to be the best thing since sliced bread. the computer could match criminal suspects up with a nation-wide database, and identify common characteristics, fingerprints, known aliases, etc. it was a costly flop-and the system never worked as planned. So why are these huge databases so difficult to use? Are most crimes still solved the "old fashioned way?
"?
I believe H.O.L.M.E.S. the UK police criminal database has been quite succesfull, though I can’t find any details as to how many investigations it has been used in.
IMO, most crimes are solved through good detective work like witness interviewing and suspect interviewing. Physical evidence and CSI work is also very important, but not as important as CSI, the show, likes to pretend.
Some computer databases are better than others. CODIS, the national DNA database, is extremely good at providing solid evidence for unsolved sexual assault cases. I have had several cases where the unidentified suspect was sent to prison on another case. As he is sent to the penn, his dna is uploaded into CODIS and whammo, you have a dna match.
These programs are not perfect and never will be, but they are very useful tools. Nothing, however, can beat a solid and sharp detective working the case. BTW, I was a prosecutor for 5 years and I now do criminal defense.
DNA and fingerprint databases have proven to be very useful investigative tools over the years.
OTOH, a firearms/ballistic evidence database would be a disaster for reasons obvious to anyone who stops to think about it for more than 30 seconds (hint: A particular firearm’s extractor leaves specific marks on ejected casings. I can procure a replacement extractor for my SA1911A1 for less than $2.00 or I could just bung the thing with a hammer a couple of times and reinstall it. Now that mark is different or non-existent).
Predictive systems will need years of testing before they are dependable. Even then, I think we’ll find them to be of limited usefulness to police but possibly of greater utility to insurance underwriters.
AFAIK, the fingerprint database doesn’t give you an exact match. It will pull out the 10 or 20 nearest matches and then the final matching is done by a person.
This might be a tangent, but how many crimes actually yield a useable fingerprint but no suspect? I would venture a guess that such a database is only useful when you have a finger to identify(possibly attached to a suspect who is using an alias and refuses to give a real name). Without an actual finger where a finger print can be read in a controlled environment, I seriously doubt you would be able to lift a print off an object and have enough to be able to find something in a database. Unless I’m totally wrong, that’s something I’d refer to as Miracle-turned-Mundane-by-TV.
Anecdotal Evidence: Three cars were broken into in fron of my gym at 2AM. Mine included, my stereo was roughly ripped out in its entirety with pieces of dash broken off, not a clean job. The cops came, and appeared distressed about such happenings in their little suburban town(Mountain View, CA), which I found to be quite odd, since I’d think this sort of thing happens all the time. Maybe not three cars at once, but still. They decided to dust all three cars for prints, door handles, dashes, steering wheels, seats, etc. The entire process took at least an hour, maybe more. They didn’t lift a single print, belonging to criminals OR drivers. I asked if the criminals wiped everything down, and was promptly informed that “Probably not, it’s just next to impossible to lift even a useable partial print unless the person touches the object at just the right angle/pressure”.
So why the databases?
I work for a data warehouse company that makes big databases. We ship our systems in increments of about one terabyte with a single terabyte being a tiny, development system. Petabyte systems are on the way. Large databases aren’t the problem, it’s the large and complex applications that make the data useful. The application can only be as good as the developer’s understanding of the system. We’re dealing with a problem right now that should be relatively simple and obvious, how to determine if a shipper has loaded a trailer with as many parts as will fit. We should have been further along but we are still arguing about the many different, ambiguous definitions the customer has given us. As with fingerprint matching our brains can be easily programmed to do the task however putting that same expertise into a procedural program is a bit more difficult.
groman makes a good point about how usable prints are rarely found at a crime scene. I think people have been brainwashed by CSI type show and every detective/cop show since Holmes into thinking that crime scenes are crawling with forensic evidence. Now the reason for the database is simple. Fingerprints are the only truly unique identifier that can be retrieved with simple equipment. DNA, iris and retinas may come into it but there is a huge legacy of useful fingerprints already in use.
Having darn near single-handedly coded a full-scale secure database and it’s access application–I just want to throw in that the issue is not that software is a problem. Getting a working and secure database which does everything you want and is easily used and every bit as useful as the data within it can provide is not an issue.
My suspicion would be that things proceeded as:
FBI told to get up-to-date on all that computor stuff by congress.
Some FBI guy somehow gets hooked up with an outside contracting firm.
They request of the firm, “We need a criminal database application. When will it be done? By the way, we will pay you $100,000,000 when you are done.”
“Why we’ll have your handy dandy criminal database application done in no time sir!” 
“Perfect. Congress has given us 2 months. So we will be expecting a demo in at least a month.”
!!! :eek:
Add to this probably insufficient communication in both ways–management at the firm unable to say “no” (which sucks…my god but this ruins things), and the FBI unable to define any clearer what they want than “One o’ them criminal database thingies! I mean don’t ask me, I’m not Mister Computer Head!”
And then of course, the lovely bit where a fully working and bug free is finally presented 5 months overdue (just simply because it can’t be done faster than that), the FBI guys look at it and realise that it looks nothing like what they expected; they haven’t the foggiest how to do anything with it; and even though it comes with a perfectly fine manual. The next day, and for every day following for the next several months “small requests” are sent that must be implemented “now” or no money.
90% of these will obviously serve no purpose and yet require stupid changes and necessarily poor coding to allow to happen. The other 10% will be stuff that really was necessary for the FBI–simply that there was no way for the people at the coding firm to know these items since they had received 0 input or interest from the FBI when trying to design the thing before building it. This might force several rewrites of various sections. And just in general, everything gets minorly hacked and rehacked multiple times so that the code os all torn up and no one is certain how it works or what it is supposed to do.
And of course since you are still trying to add in stuff. fix stuff, verify all this, minimize the amount of new code because you had something that worked and darnit! and such, all at the same time–with new requests coming in every day.
So you end up with the situation here.
In my case, I made it all work in the end with little to no errors (though the code was fowl)–but I was the only coder so there was no chance of coming up to a dead end where everyones individual bits just stop working together because everythings gone to hell and too much has been randomly hacked that fundamental things have been broken in obscure ways in unkown places.
While as in my case I could just say “Don’t touch! I am still in the middle of this bit and I don’t care about the next bit! Put it on the list!” and they just have to give in.
A word of advice, if you ever have problems with outside-sourced code, 1) Make sure to cut out any and all layers of management or you’ll get BS on the current state of the code or progress towards completion, 2) Actively participate in the design of the UI and feature list of the application you want and 3) Don’t let any coding take place until you are certain that everyone–particularly the people designing and building it–knows the design and requirments.
True, stuff will still need to be changed and added after you have gotten it into your hands–but in general the code will already be optimized for the kind of structure you requested, so more requests in a similar vein will generally be no problem to pop in there cleanly.
…Which inspired me to check out the page of the company after several months ignoring.
Appears that they are up for an award for the bugger and that they have extended their operation by at least a full digit! Not certain if I am happy. Wow.
(Me wants the award!)
THIS OFF-TOPIC BROADCAST HAS BEEN BROUGHT TO YOU BY LACK OF SLEEP!