That approach may be easier to defeat nowadays with DNNs - they would be extremely good at getting the noise out of the signal, so I am not sure it's enough to counter any nation state capabilities who really want to know what you are doing.
Why do you think so? If the patterns are relatively predictable and constant (which would be the case if they are programmed to issue some specific noise on a regular basis), Machine Learning will have very good accuracy and distinguishing what is signal vs what is noise. Much better and faster than humans, that's for sure.
