You'd need to differentiate between sources - you'd want to capture every signal, then sort into buckets by frequency, by regular timing, and so forth - if a device is sending a burst every 5 seconds, then you can grab every 5 second occurrence of a signal at that frequency and make a reasonable assumption that all that data is from the same radio.

You can filter for all the frequencies that show up regularly, then you differentiate by signal strength - group occurrences of the same frequency into similar dB buckets, then correlate the changes based on new fixed positions within the car, and run some calculations on changes in signal strength to obtain a dB to distance calculation. The strength to distance calculation can be estimated by making some assumptions about the type of radio you're looking for - a simple cellular module is going to be different than a WiFi repeater, or a wireless fob, or a bluetooth tracker.

From the fixed points within your car, you can tie one end of a piece of yarn to where the sensor was affixed, and the length of the yarn should correlate to your dB to distance estimate for that position, and with 2-3 or even 4-5 threads you'd be able to group their loose ends together to get a rough physical indication of exactly where the radio transmission is coming from.

The grouping won't be exact, but it'll literally point in the right direction, and if the threads are too long, or pointing to something buried in the chassis or whatnot, then you can reduce the lengths of your yarns by the same percentage of reduction and they'll be "pointing" at wherever the radio source is.

You're going to get a general location, like "under the dashboard" or "in the glovebox" or "somewhere under the spare in the trunk", not a millimeter precise location. You could probably vibecode a way of processing the data in a browser, and use a bunch of splats and AI modeling of your car and so forth to get a very precise and useful pinpoint of a device with a fancy UI, but you can just use a spreadsheet and text files of logged signal records, the process isn't super difficult.