I don't see what this has to do with your comment's parent. Also, there is no such thing as "not even the NSA should be capable of decrypting it" without also

I don't see what this has to do with your comment's parent. Also, there is no such thing as "not even the NSA should be capable of decrypting it" without also mentioning a timeline. Everything is theoretically decryptable in the future (by the NSA or by non-state actors) assuming computing power grows as time goes on. The supercomputer currently in my pocket is capable of more computing power than all of the digital computers in the world combined the year the NSA was first created. If quantum computing turns out to break all traditional symmetric and asymmetric encryption (as some many suspect is possible), it likely doesn't matter how many bits your encryption scheme uses -- it will eventually get broken if that's the desire of your adversary. Also, I see a few potential problems with your scheme. What happens if the HQ gets hacked and all private keys and public keys are leaked (say, by an adversarial nation-state)? The device becomes useless from a security standpoint until a software/firmware upgrade occurs, and then there's no guarantee that HQ didn't have an advanced persistent threat installed. By opening a hidden "hero" account at all, you are introducing additional attack surface area that doesn't need to exist. There are some devices that simply should not have backdoors. The more we allow companies to backdoor products and not pay (criminal punishments, civil punishments, customer/public/press backlash), the more we will end up suffering from insecure systems in the future.