@meshy @nik They're leaky as hell and getting worse, and there's no hardware on your machine that they won't gladly hand to whatever ad server asks for it. They operate on the premise that they're the only thing running on your machine, so they eat all of your memory, video memory, CPU, etc. Chrome downloads executable blobs and *is* spyware, Firefox is a crappier version of Chrome that gets its money from Google. (Linksys made all right gear until Cisco bought them to turn them into a mechanism for selling coupons for Cisco gear.)
Like, think about geolocation. It's not just based on your IP any more, not even GPS. Google's stupid map cars (not to be confused with
https://www.gnu.org/software/emacs/manual/html_node/eintr/mapcar.html ) were dumping wifi traffic and they apologized and were super sorry they *promise*, but that was seed data for capturing WiFi MAC addresses, devices and WAPs. (The stink the EU kicked up over the contents of the packets was a distraction at best; Google wanted the metadata.¹) They were exposing this to the public until Samy Kamkar (really cool fuckin' guy) built a little web UI that let you type a MAC address and see where that router was on a map. If the signal level from a known WAP is $x and you see a new WAP and the signal is $y, you can tell how far you are from both. If you move, you've got another data point, and that's enough to triangulate the new WAP. What if you're near no known WAPs, though? GPS lock, triangulate the new WAP. And not just WAPs, MAC addresses of devices fly all over the network. An Android device *touches* your network, and Google knows who's there and where they are (and the manufacturer of the device, and since MACs are often serial, usually the *type* of device). And the WiFi antenna isn't the only antenna: all these Bluetooth devices, all over. Bluetooth interface on your stereo? If there's ever been an Android device on your network, Google knows where it is, so even without a GPS lock, it can geolocate any Android device that saw a packet from the stereo, or any device running ChromeOS, probably any device running Chrome. No hardware Bluetooth killswitch on your iOS phone? Google knows where it is. Grocery stories already use WiFi and Bluetooth triangulation to figure out where people are, how much time they spend in what place. (Even if there's no WiFi, your phone will periodically send a beacon to try to find known WAPs; same with the Bluetooth hardware looking for your headphones. It's technically illegal to listen in on the cell phone bands, so that's not commercialized yet, but maybe the 5G band is different and maybe that's why businesses love 5G.) Someone with an Android phone walks by your house, Google knows who lives there and where you shop. Ever had a one-night stand with someone that had an Android phone? Google knows if you stayed behind to cuddle. The only reason they haven't counted thrusts² yet is that nobody's thought to do it yet, but they've got the data if they want it.
So now the browser wants to fondle your Universal Serial Bus:
https://developer.mozilla.org/en-US/docs/Web/API/USB . Why would they let shit like that in?
And of course, they can offload the heavy vector math (the bread and butter of machine learning) onto your system:
https://en.wikipedia.org/wiki/WebGPU . Of course, Google will mark a site as malicious if there is a Monero miner (not very often when there's a malicious advertiser), and Firefox takes orders from Google on that (hop into about:config and have a look at the variables that start with browser.safebrowsing, urlclassifier, and geo.wifi), so being on their shit list is already 71% of the web.
When browsers started executing code, it was supposed to be in a sandbox.
That's just part of it. Browsers are terrible now. You know who runs the W3C? It's public:
https://www.w3.org/Consortium/Member/List . Of course, it's alphabetized so "Whose checks have the most zeroes?" is not obvious, but scroll through the list. Adobe, AirBNB, Alibaba, Amazon, American Express, Apple, AT&T, Baidu, Brightcove (ads, if you don't know), Ericsson, Facebook, Google, HM Government ("HM" as in "her majesty's", i.e., the UK government), Huawei, IBM, Intel, the J. Paul Getty Trust (this is more sinister than it sounds if you're unfamiliar), Mastercard, Microsoft, Netflix, OpenX (ads), Quantcast (ads), Salesforce, Scribd, Siemens, Softbank, Stripe, Telstra, Tencent, NYT, Washington Post, Viacom, Disney, Xiaomi, Yahoo, Zoom. Plenty of others, ad companies, etc. CERN's still technically on it. So this is how you get the DNT header and then everyone ignores it and now there are cookie windows.
I don't have to worry about any of that if they don't have access to a real computer with real hardware, and I don't wanna worry about any of that.
¹ And I know this, because I've worked at the companies that do this shit and build this kind of thing and why they build it. Look at
https://en.wikipedia.org/wiki/GOOG-411 . It looks like, for three years, Google ran a free 411 service. 800 numbers cost money: the service pays the phone company for each incoming call. Why would they do such a thing? I know, because AT&T built the same kind of thing to do the same kind of thing: big bucket of training data. We had people speaking location names in every accent in the country. (In this case, anyone could know it about GOOG-411, because it's also on the Wikipedia page.)
² Speaking of location data, that "Give permission for your location data?" prompt is complete horseshit if you've got an accelerometer. See attached: you can get location data with accelerometer and having map data. (Ask OSM if you want it for free.) Bonus attachment: the accelerometers can also be used to listen to audio. The companies doing browsers, phones, operating systems, etc., are all building backdoors and leaving them open when they walk through because we let them.
motion_sensor-based_privacy_attack_on_smartphones.pdfinferring_user_routes_and_locations_using_zero-permission_mobile_sensors.pdf