According to the article, current adblockers like Ublock Origin aren't searching for keywords in the page elements for ads, they're just using human maintained lookup lists?
That's very surprising. I would have thought that would be the first avenue of attack for adblockers.
Also, who wants to help me retrofit this thing to actually block ads?
How do you know what the development environment 'felt' like? Are you Roger Dingledine or Nick Mathewson?
And I doubt that you will find any evidence that it was used before being made public. Using TOR before it was public would be like screaming, "HEY I'M HIDING SOMETHING! AND I'M US MILITARY OR INTELLIGENCE!". The whole point of releasing it was to gather a userbase. Otherwise, TOR wouldn't be very anonymous at all.
Noise. Deep learning looks for trends in data. If there's a bunch of noise (extraneous random data), it's very difficult to put together trends.
For example, lets say you want to throw off Google's targeted advertising algorithm. Let's pretend you want to buy a stapler. Right now, if you go visit a bunch of sites that sell staplers, or a stapler manufacturer, then you'll probably start getting ads for staplers. In order to counter this, you could visit a bunch of random sites while mixing your stapler research in. This way, it's difficult to spot the trend.
This is a gross simplification, because a smart algorithm is going to see that you've visited a bunch of random, non-correlated sites, and eight stapler websites. So the noise that you're hiding in needs to be specifically crafted. For example, you still want to do stapler research. But you know now that you need to be even sneakier. So you visit twenty sites on subject a, twenty on b, etc, then you do all of your stapler research in a single 'window', and then continue onto more random subjects. Thus it's more difficult to figure what you were actually researching. You could probably write a script to do this pretty easily.
I'd like to tack onto number 3 though:
3a) Don't try to learn everything. Being a jack of all trades and a master of none isn't the solution. Be able to write good code / thoroughly utilize the technologies and frameworks that you DO know. Then progress from there.
I'm not sure where this narrative is coming from. TOR was developed as an anonymous network by the US Naval Research Lab. It was designed for use by military and intelligence. TOR was never just some academic experiment.
TOR is still a valid tool. No, it wasn't designed to foil NSA level surveillance, because it was built by the US. But this vulnerability isn't even related to TOR, it has to do with the TOR Browser.
The Snowden leaks contain slides where the NSA clearly laments the use of TOR, so saying that it never has been trustworthy is simply not true.
Well done video. My only complaint is the stack being inverted a few times throughout. Saying it grows down and then showing it growing up might confuse newbies, who I'm guessing are the target audience for this video.
Other than that, your explainations are clear and concise, which is rare for this topic.