So I started methodically learning cybersecurity. I ended up writing this comment about a year or two in:
https://news.ycombinator.com/item?id=12788910
> It's surprising how bad cyber security is, but so much of it is right there in the pages of this [U Waterloo textbook]. It's like finding out you can buy a Patriot missile for $250 and some spare time in the evenings.
Over the past two years I think the world is starting to understand where we are headed, but there are no easy solutions. Many of the problems we have are the same ones we've had since the dawn of the NSA. Computers need operating systems. Operating systems are really fucking hard to make completely secure, let alone completely secure and borderline useable. If you think $500k for 0days is high you ain't seen nothing yet. When autonomous systems run everything or when greater numbers of Wall St traders start realizing what they can do with an 0day and some outlying put options we're going to see 0days worth tens of millions as the arms race heats up.
The problem with cybersecurity tools (and AI / autonomous systems) is that it's all dual use. The same tools you probe your own server with are the same ones you can probe others with. Worse—we can't even control the export of attack tools because it's all essentially just data. You can't stop people from memorizing code snippets or facts.
Even so, we need to instil fear into people in the West. We need to limit who they're legally allowed to sell the vulns to. Allied states: Yes. Defence interested parties: Yes. Some cybergang: Fuck no. We need to deny travel visas to the direct family members of other individuals in unaligned states that sell 0days to the worst actors.
It won't stop malactors from getting these bugs, but it will make it significantly more expensive for them to do so, and at the end of the day war and crime are economic concerns as much as they are political.