In the last days I was looking on Tutanota, ProtonMail, LavaBoom, and others webmails that promise to respect users privacy. For all those, the one I most liked was Tutanota.
I did few tests with this one, sending e-mail for some account outside it, what I saw was: It encrypts a symmetric key with a chosen password, and sends a link for the recipient e-mail address. Ok, but if I change the password for another message, it keeps decrypting the old ones with the new password. I asked them and they told me that 'indirections' are used.
My question is: let's suppose we have hybrid encryption on all this e-mails (when communicating between them, for example: tutanota sender/tutanota recipient), it's encrypted with AES random key, and the random key is encrypted with the recipienr's RSA public key.
But the entire scheme still depending the same RSA key, time after time, even when user change his password. If this key, sometime gets compromised, the user cannot fix his problem just changing his password, right? If things are like this, I believe that this is a implementation problem (maybe just a concept problem), because if the system grows up, and have 100 millions users, and become a standard, in some years RSA 1024/2048, what they should use right now, will become overdue and easily crackable. The millions of users will have millions of data encrypted with overdue encryption, and they will need to re-encrypt everything.
I'm missing some point, or this will happen? I know that every encryption will become overdue in the future as the processing power increases. But most of the systems, can easily update their encryption and make the new algorithm the standard for the entire system, this is not the case of a webmail system.