I think this relies on the old argument that anything connected to the internet is potentially insecure. While it might have some truth, practically we all do very sensitive stuff securely while connected to the internet. The risk is there, always, but you put all the measures to mitigate it and even prevent it.

The idea that a malware could be on a phone “altering things automatically” feels like a 90s FUD cliche. If an online voting system existed, it won't be like a poll that you see on Twitter, for instance; it will be far more involved. For example, we can have blockchain as the network, and not just transparent to all, but even after you vote you can still check your vote and see if it was potentially altered, and a proper electronic chain of custody can also ensure that the vote was counted per the process, and all of that is visible to anyone who would like to check and even count ALL the votes yourself, again, just like how transparent blockchain is.

And saying paper voting is more secure isn't true at all, because these votes will be counted electronically at some point, either by a machine or just a simple Excel sheet, opening the same risks as the previous one except here, if it would happen, you will never know and you as a voter can't trace the vote from when you voted all the way until it was counted. The voting process should be designed in a way with zero trust in mind, just like how secure systems are designed now, like storage, encryption, vpn, etc., and voting should too.

I personally believe that we can build a very secure, robust, and trustworthy system that can be used for voting online, but I think no one wants that for all sorts of political purposes, either by actually altering the results that could go unnoticed, or at least keeping the window open to blame the results on a faulty system.