3 ms·

How would the browser know which to block? `onClick(){fetchAllForumTopics();}` vs `onClick(){sendTrackingData();}`. Both are simple ajax calls. Even if you so

How would the browser know which to block? `onClick(){fetchAllForumTopics();}` vs `onClick(){sendTrackingData();}`. Both are simple ajax calls. Even if you somehow managed to know which calls to block, you could simply hack it by calling a jpg with extra parameters and then have the server save those: `onClick(){body.css({backgroundImage: 'http :// mysite.com/ images/ logo.png?x=123&y=512&elementName=BankAccountNumber'});}`

notyourwork 8y ago
Even ignoring that knowing where you click is as simple as knowing what feature you use. If you login then I know you clicked login, if you perform a search then I know you clicked on the search input. Usage of an element tells you where a user clicks. It misses things like clicking on whitespace or where your cursor traverses.
- sailfast 8y ago
  Right - that wouldn't be a perfect privacy solution but it would likely prevent significant biometric data gathering and not be too huge a burden on users / adoption.
sailfast 8y ago
I wasn't thinking of blocking specific function calls so much as that data not being available to the browser at all to send in the first place, providing the ability to selectively block mouse x,y coordinates tracking, etc before the code on the page was executed. Again, this may not be possible without breaking the web given the structure of the browser and what pages need to render at a low level. Perhaps client-side middleware that only passes user data when an action is performed rather than constantly streaming coordinate and interaction data all the time?
- Kalium 8y ago
  This is a great idea! It's always a good idea to control carefully what information is available to a system you might not always be able to trust completely. It's especially important to be sure you're not leaking info through a channel you didn't consider, which at times means that an allowlisting approach is preferable to a blocklisting approach. Your heart is in the right place. Keep iterating!
CuriousSkeptic 8y ago
I’m thinking broadcast models of publication is inherently less privacy invasive. So a browser only functioning over a federated content addressable web fetching content from trusted cache nodes could help isolate the problem to only vet information intentionally meant to be communicated leaving pure consumption reasonably private.

notyourwork 8y ago

Even ignoring that knowing where you click is as simple as knowing what feature you use. If you login then I know you clicked login, if you perform a search then I know you clicked on the search input. Usage of an element tells you where a user clicks. It misses things like clicking on whitespace or where your cursor traverses.

sailfast 8y ago
Right - that wouldn't be a perfect privacy solution but it would likely prevent significant biometric data gathering and not be too huge a burden on users / adoption.

sailfast 8y ago

I wasn't thinking of blocking specific function calls so much as that data not being available to the browser at all to send in the first place, providing the ability to selectively block mouse x,y coordinates tracking, etc before the code on the page was executed. Again, this may not be possible without breaking the web given the structure of the browser and what pages need to render at a low level. Perhaps client-side middleware that only passes user data when an action is performed rather than constantly streaming coordinate and interaction data all the time?

Kalium 8y ago
This is a great idea! It's always a good idea to control carefully what information is available to a system you might not always be able to trust completely. It's especially important to be sure you're not leaking info through a channel you didn't consider, which at times means that an allowlisting approach is preferable to a blocklisting approach. Your heart is in the right place. Keep iterating!

CuriousSkeptic 8y ago

I’m thinking broadcast models of publication is inherently less privacy invasive. So a browser only functioning over a federated content addressable web fetching content from trusted cache nodes could help isolate the problem to only vet information intentionally meant to be communicated leaving pure consumption reasonably private.