Learning to Fly
- May 28, 2017
- Reaction score
- Mojave Desert, CA, USA
While I think this is an interesting discussion I don't want to overemphasis it. I think it is neat to brainstorm on it, but I think it may be too large a problem to be something real anytime soon.With the caveat that I haven’t tested it yet, I’m not so sure that is the case.
If I record encrypted police recordings and non-police recordings and then feed them to RAI to train on, I would bet money that the AI would pick up on “tells” or patterns in the police recordings that don’t exist in the other recordings. For example, if they always start their transmissions with a certain identifier for dispatch, etc.
These are things that we wouldn’t need to break the encryption to see - the AI doesn’t know what needs to be said during the transmission to tell the difference, it just needs to see something different between police and non-police transmissions.
But with that said, you may want to take a closer look at the format of the data used in most such systems. Keep in mind I am NOT an expert on these systems, I do not work with them professionally, I have just touched on them in passing, but what follows is my understanding of them from my observations. I would be glad to be wrong, and if I am...yaaahhh! Maybe it would work as you are thinking and that would be very cool.
If you record encrypted police and encrypted non-police recordings the data format will be indistinguishable until you decrypt. The data outside the encrypted information, of course, will be usable. Something like NAC, unit ID, and TGID for P25 transmissions. But unless you have a database of known stations (probably correlated with specific frequencies) that is not going to do you any good, because the non-police transmissions will also contain the same classes (but different specific data) of information. None of that data is specific to police transmissions, it is specific to the network being received, which may be specific to agency or municipality.
What I am basically saying is that once they are digitized and encrypted there are no repeatable differences between police and non-police uses of the same encrypted digital system. The only differences are the words being said, and since that is digitized and encrypted, and not all police agencies use the same formats, that will be a problem.
I think without using frequency as a discriminator it is going to be a very tough nut to crack. Not undoable, but pretty tough. Fortunately, freq databases are readily available.
One way to use encrypted LEO transmissions is to watch only the repeater or system input freqs, vs the output freqs. The repeater outputs are going to have a large footprint, covering wide areas. But the mobile and hand held units, on the input freq, are typically going to have much more limited areas of coverage. If you hear one it is not too far away, depending on terrain and situation maybe only a couple of miles or less. Make the system trigger only on very strong signals, or not worry about developing highly sensitive receivers and antennas, and you could trigger only on things inside shorter ranges. Again, for this to work you need to know the input freqs, but like I said, the databases are widely available.