Good luck with all that coding. I have four years now of running my own email server. Zero hacks. I keep the attack surface to a minimum. Less is more.
One thing you don't want to do is write your own code. This stuff is always way harder than you think. Worse yet you run alpha generation code because you are the only one using it. All software has bugs. What you need is a mass of users flogging the code and finding the bugs.
Now if you do use a browser, you have to deal with leaks, bugs, possible process interaction if more than one tab is open, and possibly browser extensions hacks if extensions are used.
Count me out.
And did you miss the part where I was hacked via RoundCube?
Original Message From: sebastian@sebbe.eu Sent: October 25, 2020 9:47 PM To: dovecot@dovecot.org Reply-to: dovecot@dovecot.org Subject: SV: Looking for a guide to collect all e-mail from the ISP mail server
"Never use a browser for email."
I don't agree. In fact, using a browser for email or atleast initial setup, is actually more secure. This because SMTP/IMAP clients normally don't support 2FA, so you would have to "hack" a solution to enable 2FA for email.
This can be made in 2 ways: Either, you have a full fledged email setup. Whats important, is, to prevent auth-bypass holes, you remove the authentication in RoundCube or whatever webmail you use, and instead use a reverse-proxy or firewall authentication instead. Thus an unauthenticated user doesn't even touch RoundCube/webmail at all, but must authenticate at a prior stage.
The second way, is to not have webmail at all, but instead have a authentication gateway in browser, where you must auth with 2FA and captcha. The only purpose of this gateway, is to authenticate users with 2FA before their IP is whitelisted.
After this, you simply have a script, that upon valid login (with 2FA) in either webmail or auth gateway, you set the authorized IP of the user to this. Whats happen then, is that each account will have an authorized IP attached (you could limit it to the /24 to cater for mobile clients), and then login to that account, will only be accepted from that authorized IP.
This then allows SMTP/IMAP usage from that IP. If you want to go even more secure, you could restrict the firewall to the list of all IPs that all users have dynamically, and then in the SMTP/IMAP server, lock down auth to the authorized IP of that particular user account only.
Its very important, that upon authing with a incorrect IP, that the server responds in the same way as a invalid password was specified, in this way, if someone attempts to bruteforce the password, they will "miss" the correct password, if the server does not react differently to a correct password but invalid IP. Thus bots that bruteforce will not gain any success.
All this can be combined with permanent whitelists and geoIP whitelists, to avoid users having to authenticate with 2FA for "trusted" locations. One example would be to have the local office as permanent whitelist, and also have it that any IP in the user's "home country" is permanently whitelisted for his account once the user authenticates with 2FA.
Other IPs outside his home country, is then only whitelisted once, next 2FA login, the old whitelist is simply deleted.