I understand your concern. But what if the system is built on top of a secure and reliable blockchain network like Bitcoin? Then it would be almost impossible to inject malware and manipulate votes.
The malware problem is actually at another place ... the device of the voter. People could be tricked to install a software which can access the keys for voting and vote for them, or falsifies their vote. Take into account that blockchain voting would be an option which has to be accessible even for non-technical and poor people who are vulnerable to these attacks.
Of course you can try to make everything safe with biometric access, as was already discussed. But first, this introduces complexity and additional vulnerabilities (the biometric data can be stolen, for example). And second, this would perhaps make voting as secure as banking, but it's not enough. Voting has normally higher standards. Once the vote result is official, you can't "chargeback" the votes as you can sometimes with erroneusly transferred or scammed funds when your banking app was hacked or tricked. If you allow "changing votes" after the election event, like chargebacks in banking, I imagine then this would become a total mess which would be exploited by all parties to gain advantages.
The most secure way would perhaps to offer a dedicated device for voting. Or distributing an official app in an offline way (e.g. where you can get the download link only at a government dependency) But even if the technical side is totally safe, not only but principally in countries where corruption is rampant, party officials could offer you "benefits" (like money, free lunch or so) if you vote "with them" for their party in their party office - i.e. enforcing secrecy would be an enormous problem.
Some of these problems of course also occur with postal voting, and countries who accept that method are accepting this risk. But online voting, including blockchain voting, even then introduces more vulnerabilities. Estonia and some other countries do already apply online voting and they seemingly accept the risk. However, in Estonia's case where the database operator is probably trustable and the country is relatively small and "controllable", blockchain technology would not add much safety. So I believe the only use case I really would approve is what I wrote in my previous post - blockchain as a backup for a semi-electronic system which can be also manually controlled, and where there are minor trust problems in the "voting chain", as in most South American countries for example.