By not executing code that exploits that vulnerability, or preventing it from executing in a way that causes harm.
For example, the Firefox 57.0.4 patch prevents (or seriously mitigates) javascript from being used as an attack vector: https://blog.mozilla.org/security/2018/01/03/mitigations-landing-new-class-timing-attack/
There is a crapton of hardware errata in any modern processor, and firmware or software mitigations are common practice. Eventually, once essentially everything is patched, it will be exceedingly rare to find an attack vector even on vulnerable hardware, and people will stop trying to take advantage of the vulnerability because the number of viable targets will be minuscule.
Meltdown and Spectre are pretty serious issues, but it's not the end of the world.
Anyway, the AV patches mention aren't so AV can protect against Meltdown and Specter (they can't, except by detecting and quarantining malware/viruses that use it), it's so the Windows mitigations don't brick the AV (which essentially use rootkits that the Windows patch would break, making unpatched versions of the AV nonfunctional).
But it has to detect the code that exploits vulnerability first, which is, both judging by common sense and the original paper published by research team which discovered vulnerability is "theoretically possible but highly unlikely".
Yes, sure there are a lot of CPU (and not only CPU) "bugs" which can be exploited, rowhammer is nasty one too as an example, and some of them are not even completely mitigated especially in home PC-s, but when there is such hype around some of them it usually leads to a lot of exploits "in the wild"...
Patching AV itself to protect it from exploit makes sense though, unlike claims about protecting OS...
Also, played around with proof-of-concept code which was published along with the paper, and it is a lot of fun. For example it was easily capable of breaching guest isolation from inside VM running on rather old version of vmware player on unpatched windows.
Last edited: