Say I am at my office. I am now trying to talk to my friend with a LUI instant message application.
I have a blanket covered over my keyboard, just like Snowden, probably good enough to hide my password. However, there are two problems:
Firstly, someone might replace my keyboard with a one that look visually similar but have a slightly different sound produced for each key I pressed, so a nearby sound interceptor can intercept the keys I entered.
Secondly, my friend's and my messages are shown on the screen with totally no protection. An inexpensive camera would suffice to capture all the messages. I was thinking about implementing an encryption algorithm, communicating over my computer and me, but I think either it is too insecure or it is too complicated for my mind to handle, e.g., I cannot compute even plain RSA in my head.
How can I mitigate the sound and visual attacks?
I understand that there is no "perfect security". I hope to reduce the risk instead of eliminating them.
For problem 1, I am currently only targeting acoustic attacks. It is taken under the assumption that even if the keyboard is bugged, only acoustic attacks will be carried out.