The cable modems I've interacted with all seem to be immutable infrastructure: their state consists of 1. a signed firmware image; 2. a DOCSIS initialization packet received upon network registration (essentially equivalent to a cellular "carrier profile"); 3. a set of saved preferences; and 4. some volatile working state, like ARP tables or WPA encryption-stream key state.
#4 gets trashed on reboot; #3 shouldn't be large enough (or executable) in a way where programs can run from it; #2 gets overwritten every time the network reconnects; and so the only place to put a persistent malware vector should be #1. And #1 can only be written to using signed code (sort of like Intel CPU microcode.)
So what's the problem that can't be solved by rebooting the modem?
They may technically own the modem but you’re in your own to keep it running.