Schneier on Security

Doug Coulter • March 23, 2008 6:04 PM

I am a recently retired digital signal processing consultant, and am currently a gunsmith, so I suppose I can fill in some facts for this discussion.

But opinion/politics first:
I think the emphasis on possible misuses/abuses of this kind of system is (very sadly) completely well founded, and backed up by publicized actual occurrences of abuse. Police have, in my opinion, far too much discretionary power, and have lately begun to abuse it more and more. In my small town, this is not a very big deal, where everyone knows everyone, and abuse is not rampant -- if a cop does something bad to a good citizen here, he won't last very long in his job.

Law enforcement has a much tougher problem in a dense population where everyone is a stranger, however, and grasp at whatever seems to work -- and that's the honest, caring ones. Having lived in both sorts of places, the differences are pretty obvious and well confirmed, at least by me. In my small town, it is against the law to go into public without ID AND at least $200 in cash, else one is a vagrant. It's something the cops keep in their back pockets if they want to look one over for awhile, in lockup and have nothing else to go with at the moment. Here it is even a felony to have some kinds of consensual sex acts with one's legal spouse(!) -- also just about never used -- but it could be.

Selective enforcement is a very dangerous thing to our rights in general. It lets politicians claim they are doing something and pass some very rights-threatening laws, and in the better cases where they aren't abused, prevents the huge backlash that would occur if those laws were enforced, say, as vigorously as even speeding on the roads, or public drunkenness, which many people regularly get away with. But as they are on the books, and largely unchallenged, this could change at any instant, at essentially the sole discretion of an officer, who in effect becomes judge and jury on the spot. I thought we had separated those functions for sufficiently good reasons!

As to the DSP of detecting gunshots, it's a relative piece of cake. Gunshots are characterized by a supersonic muzzle blast, followed by the usually supersonic bullet, both of which have risetimes not created by almost anything else. All this is followed by multipath echoes from which various parameters can be determined with pretty good accuracy. Our military has looked into this for their own uses, with success.
As taxpayers, we've paid for that already.

I agree that recording gunshots or explosions is tough, but it is still very do-able if one sets up for that.
I've done it, and used a gunshot for a snare drum sound in the recording studio. There are even records available of the sounds of cannon fire for Civil war re-enactors and historians, and they sound pretty realistic. What IS difficult is reproducing them with most current audio equipment.
The SPL is just so high that human perception is in a nonlinear mode (not to mention the air itself -- it can compress to almost any pressure, but not rarify below a vacuum), so a level that wouldn't cause legally actionable damage to listeners doesn't subjectively sound the same without a lot of compensations. I've heard it done well in a movie (Terminator) with one exception -- after the first shot inside an elevator, one's hearing would be severely affected, and subsequent shots would not sound the same.

What got much more interesting for my company was when we were tasked to do a "scream detector", which has a lot of the same system-gaming problems. This was for a customer who makes PA systems for schools, airports, and other security related equipment. In a classroom, especially elementary, students scream a lot, and it means nothing except a disciplinary problem -- and a wide open invitation to abuse such a system. So, it boiled down to "how to tell a real scream from a fake one?". Imagine figuring a way to take data on that without accidentally scaring someone to death or winding up in court! And not everyone screams when suddenly frightened. Sadly, the project never got past that obstacle, and I doubt one ever will, unless someone just takes data until real events occur enough times for a significant data base to be built. Obviously fake screams are easy to get data on by comparison. At least it was entertaining to think about how to truly scare someone without backlash -- but no solution eventuated.
I personally believe the detector could be done, but not without good data. Would have been fun, but...that's a really big but.

As a gunsmith, I can say that BATF class III devices, which include machine guns and silencers are legal, but it is so much hassle with paperwork and fees that only the very wealthy can get there from here. This is a federal law, carefully crafted to avoid running hard into the 2nd amendment, but still accomplish the severe reduction in availability of such things in general. It's about as hard as getting a very high level security clearance, but you do the paying for all expenses, rather than an employer, and there is no market competition for the costs. States may pass various laws, but face a test on pre-emptive restrictions if they get taken to court on them, in which cases they mostly fail. After a very expensive and time wasting bunch of legal wrangling.

Our gun club has a semi-annual machine gun shoot, well attended by law enforcement folks, for example, where one can rent time on various legally owned full auto guns. Fun, but it's expensive! Bottom line, these things can be legal, if you have a squeaky clean record, lots of patience, and tons of money to get the permit, then another ton of bucks to buy the device. This is one reason that you almost never hear about one being used in an actual crime if the reporting is at all accurate. They are just not common at all, and so are even hard to find to steal from someone who legally has one. Not many people want this kind of thing badly enough to handle the 6 figure total price, and the extensive waiting involved. Criminals lack money and patience, one reason we've also not seen crimes with .50 cal rifles -- they cost far too much to be in any way common, and to boot are darned difficult to move and disguise due to size and weight.
Walking around with one attracts a lot of attention, even at a gun show, as I noticed just yesterday.

The movies give a very false impression of the capabilities of silencers, by the way. They only really work for subsonic bullets, and guns with no other gas leaks, which eliminates all revolvers, and semiautomatic pistols that haven't been modified to be single shot by locking the slide closed. Due to aerodynamics, the bullet velocity must be well below the speed of sound at about 1130 fps -- more like 800 fps, or it gives the loud supersonic "crack" as the air has to move a little faster than the projectile to get out of the way of it. Revolvers and pistols leak, around the cylinder-barrel gap in the former case, or by unlocking during high pressure in the latter. It is still quite a bang, in short. So the TV/Movie "zip-zip-zip" is completely fake and never encountered in real life. There was some work in WWII with subsonic rifle rounds with very heavy bullets (so they'd still have enough energy, even if slow) by resistance fighters, but one had to be very close for that to work.
You'd be far better off with a crossbow.

Even in some Scandinavian countries where silencers are legal (and encouraged for hunting!) they don't work all that well, even though there is a market and therefore design impetus for continuous improvement. It's just really hard to make a good enough acoustic low pass filter to cut the blast down much more than a few 10's of dB, at least a portable one. And considering that you may be starting essentially "off the scale", cutting the blast down to say, 140-160dB absolute doesn't by any means make things "silent", "quiet" or anything but less-immediately damaging to the user's hearing.