Paul Kocher was invited to give an invited presentation at Crypto and CHES, which was certainly deserved on the 20th anniversary of his paper that has more than 3500 citations on Google Scholar. The content of his talk ranged from tales of his work over philosophical considerations on security to an outlook to the future.
It was interesting to see how Kocher, a biologist by training, got into cryptography and managed to break implementations via side channels with rather cheap equipment, including parts from a toy electronics kit. He claimed that they could break every smart card at the time, which was of course disputed by the vendors.
In the philosophy part of the talk, the speaker brought up various analogies that I found interesting even though they did not provide direct advice. For example, he compared security to fractals that provide ever more detail the closer you look. More practically, Kocher mentioned building codes and the aviation industry. Both minimize risks by best practices that include safety margins even though these incur extra costs. However, I could not help thinking that aviation killed a fair amount of people before the standards improved.
On the outlook, Kocher did not seem particularly optimistic. The world is already complex and full of devices that regularly exhibit security problems, but it will get worse with the Internet of Things, where there will be more devices with longer life spans produced by vendors with less security knowledge while the impact of vulnerabilities will be even higher. He predicted that the security breaches will get worse for 3-5 years at least.
In terms of constructive ideas, he suggested to move the security into chips because it won't be ruined by the lower layer there. There already has been a significant move in that direction with Intel's SGX, but there are of course other approaches.