2 ms·
Welcome to the cyberwar. I think the basic problem with brakdown of static analysis tools like this one or optimisations like speculative execution is that the
by xg15 9y ago
Welcome to the cyberwar.
I think the basic problem with brakdown of static analysis tools like this one or optimisations like speculative execution is that they were built in a different time for significantly different requirements.
When the theories were developed, the main concern was to prove correctness and protect against accidents. Yes, the OS isolated processes from each other and protected memory boundaries - but this was added to keep a crashing process from corrupting other processes and bringing down the whole OS. It was not developed to contain an actively malicious process.
At some point the paradigm changed and the idea spread that an OS shouldn't just protect against misbehaving code but also against misbehaving programmers. That lead to the current practice of compartmentalizing a machine in different areas of trust and tightly restricting what a program can do - and at the same time permitting increasing amounts of "untrusted" code to run because we trust the boundaries so much.
I think the general paradigm shift has some worrying sides, but more importantly, it was never explicitely discussed and it was never discussed if the current tools of software and hardware development can even support that new paradigm.
So yeah, in short, I think we are currently discovering that our houses are in fact not built to withstand a 20kton TNT blast set off at the front door. So far this hadn't been a problem as random violent explosions used to be a rare occurence. Somehow we got ourselves in a situation where they aren't rare anymore.