Please note that most crypto systems fail due to a flaw in the implementation rather than its formal specification.
So for all intents and purposes, that:
[...] own encryption implementation, or anything..
...is pretty much as bad as that:
Of course you have to use proven concepts / algorithms and not create your own crypto protocol
But i thought this was obvious.
Just sayin'
I think we both have a different understand of the term 'implementation'.
If i write python code which uses a well known libary to implement AES encryption before sending it via the internet and the server receives it and uses the same libary to decrypt it... that's my 'own' implementation (in my terms).
This doesn't make it unsecure.
I did not intend to talk about implementing the encryption process itself from scratch.
But about implementing encryption at all (e.g. in an own small application).
By the way.. if you follow the specification properly, your cipher text won't be vulnerable to attacks.
Own implementation of the encryption process often leads to side effects which can be abused.
But in a MITM scenario, those side effects (e.g. CPU time) won't have any effect and therefore won't be an attack surface at all.
If you actually can read and actually can type, you should be able to create your own encryption algorithm (based on a given specification) without being vulnerable to MITM (only talking about MITM here).
P.s. Note that AES is just an example. Pure AES is not suitable for this case (missing replay protection etc.)
