Security through obscurity is about not revealing how the system works. Relying on computational complexity is precisely the opposite, you can understand the entire system but the unknown is well known, e.g. the length of the key and which symbols can be used to compose that key, but recovering it is prohibitively expensive. It's practically impossible even in theory it is possible.
FWIW we don't know for sure that one-way functions truly exists, or that P is different from NP. We just didn't find efficient solutions "so far". So the assumption for both what we used so far and this new work... is that our understanding of mathematics is shared and correct.
FWIW we don't know for sure that one-way functions truly exists, or that P is different from NP. We just didn't find efficient solutions "so far". So the assumption for both what we used so far and this new work... is that our understanding of mathematics is shared and correct.