I've worked on a team that reverse engineered and did security audits on a lot of commercial and consumer applications. We've seen cert pinning implemented correctly was maybe like once or twice a year by companies large enough to where their security team was larger than most software companies entire payroll.
Basically, it's not a thing that exist because it is really hard to implement properly. The threat model for being MITM'ed with cert spoofing is pretty exotic. In the end, cert pinning means your application is not working if something goes wrong with the certs, which EVERYONE at some point forgets to renew, or, worse, you CA inadvertently gets hosed.
The few times I've used one they also take a ridiculous amount of time to connect/return anything, on the order of 30s-1m.
The jacks official name for the jack is Modular connector[1]. 6P6C for RJ14, 6P2C for RJ11 and 8P8C for RJ45 (what you described as cat5 jack)
Fun fact, the modems still negotiate at 2400 or 9600 baud, because the extended negotiation times of higher-speed protocols more than negate time saved in transferring the small payload.
They even make gateways to bridge old dial-up ATMs to IP: https://atmpartmart.com/wireless-atm-modems/systech-box-ipg-...
It seems like the ATM's software might work like this: on bootup, connect to server atm.foobar.com at port xyz.
Oh, right. In that case, you'd write an MITM server. You could sneak a raspberry pi so that it goes ATM <-> RPI <-> ethernet, and then set up the RPI to broadcast all the network traffic via a wifi dongle to your laptop.
But... certificate pinning would trivially subvert that. I guess ATM manufacturers might not have done any pentests though, so perhaps they don't do cert pinning.
Assuming cert pinning, is there still a way?