You can pull off attacks like this from JavaScript by repeatedly recording the time and training a machine learning model on traces of instruction throughput over time, which my group did in a recent paper: https://jackcook.github.io/bigger-fish/
How would you ensure that the user loads your malicious script, and has a running web worker for it?
I see that you trained it on 100 websites. Would you need to retrain for every new version deployed or different paths with varying content?
If your intention is to detect sensitive website accesses, wouldn’t you need those websites to be public to train the model first? I’m not convinced that detecting porn access is particularly malicious, but I acknowledge that it is illegal in some places.
We evaluated on 100 websites as a proof of concept, but we also included experiments in an "open world" setup where the classifier has to predict whether the activity is from one of 100 sensitive websites, or whether it's none of them, and found that it's still very accurate in that more realistic setup. You would need to retrain to identify more websites outside of your set of 100.
The websites would need to be public, which is basically the same limitation as hertzbleed, since they need to know what they're looking for in order to identify an activity. Some use cases with this limitation aren't too hard to imagine: maybe you're in a country that bans access to major Western news sites but you're evading censorship with a VPN.
I’m not saying your paper is technically wrong, just practically infeasible.
Right now, you’ve chosen very specific websites. Have you explored if there is a correlation between specific scripts (react, jquery, etc) and whether websites with similar setups cannot be differentiated? I was also curious about content/non-homepage paths. Your conclusion seems to be that interrupts/etc are the primary indicators, so I suspect there’s a connection.
Edit:
In my experience, large websites and most web apps don’t use CDNJS/etc, but bundle their code - this would make injecting your script much harder without a supply chain attack.
On second thought, given CORS I think this attack is actually impossible. How would your embedded script communicate your findings with your server? You would need to control the originating domain itself…
We did get a comment about this in our rebuttal but didn't end up including it in our final paper -- we found that we distinguished sites with the same frameworks (such as react, angular, and jquery) at the same accuracy at sites that used different frameworks.
We didn't do much research into content/non-homepage paths but it's a good area for future research. I would suspect it'll still do pretty well.
And yes, we concluded that the source came from interrupts (in Table 3 of our paper you can see we ran an experiment with frequency scaling turned off), which does make me question the practicality of hertzbleed. I wouldn't doubt it can be exploited somehow though.
On second thought, given CORS I think this attack is actually impossible. How would your embedded script communicate your findings with your server? You would need to control the originating domain itself…