4 ms·

Yeah... Can't really trust Tor these days can you? With unique members taking over multiple relays and exit points they can now relate where you're coming from

Yeah... Can't really trust Tor these days can you? With unique members taking over multiple relays and exit points they can now relate where you're coming from and where you're exiting... Really doesn't matter anymore how extensive it is the browser fingerprinting protection feature or the access you do through Tor. I believe we need a new type of Tor...

ravenstine 5y ago
https://geti2p.net https://geti2p.net https://yggdrasil-network.github.io/ https://yggdrasil-network.github.io/ Granted, neither of those are designed as proxies to the clearnet... and maybe that's one of the bad things about Tor besides its its history with DARPA.
southerntofu 5y ago
This was always accounted for in Tor's threat model. That's why you take many different circuits depending on what location you're trying to reach. That's also why for the highest security needs VPN+Tor is a recommendation. Although to be fair i personally believe if an actor is powerful enough to perform traffic analysis across the Tor network, they're probably powerful enough to correlate your computer activity with the VPN<->tor link. Networks that resist metadata analysis are called mixnets and there's some interesting research about them. The downsides are you add latency (because each hop needs to randomize sleep) so bidirectional sessions like TCP is unthinkable. So nothing as usable as Tor Browser for checking your webmail or reading a blog.
- summm 5y ago
  Or, instead of TCP, content-based storage as in Freenet.
  - southerntofu 5y ago
    Yes P2P CAS (torrent/freenet/ipfs) sounds very interesting over mixnets. Do you have links to specific implementations?
    - summm 5y ago
      Freenet manages to implement some kind of anonymity without mixnets, relying on inserting/retrieving storage blocks sacrificing latency. In my opinion that's the better approach, because if you separate storage and communication, like ipfs-over-tor, you'll never get anything with proper anonymity and decent performance. Beyond that, I've heard of xolotl related to gnunet, but I am not sure to what extent it is actually implemented and/or working, as usual with gnunet.
      - southerntofu 5y ago
        Are you aware of modern Freenet reimplementations, by any chance?
  - goodpoint 5y ago
    content-based storage is not "instead" of TCP. It's a different solution for a different requirement.
    - southerntofu 5y ago
      correct, content-addressed storage is often implemented on top of TCP. I wrongly used TCP to generically mean "location-addressed".

ravenstine 5y ago

https://geti2p.net https://geti2p.net https://yggdrasil-network.github.io/ https://yggdrasil-network.github.io/ Granted, neither of those are designed as proxies to the clearnet... and maybe that's one of the bad things about Tor besides its its history with DARPA.

southerntofu 5y ago

This was always accounted for in Tor's threat model. That's why you take many different circuits depending on what location you're trying to reach. That's also why for the highest security needs VPN+Tor is a recommendation. Although to be fair i personally believe if an actor is powerful enough to perform traffic analysis across the Tor network, they're probably powerful enough to correlate your computer activity with the VPN<->tor link. Networks that resist metadata analysis are called mixnets and there's some interesting research about them. The downsides are you add latency (because each hop needs to randomize sleep) so bidirectional sessions like TCP is unthinkable. So nothing as usable as Tor Browser for checking your webmail or reading a blog.

summm 5y ago
Or, instead of TCP, content-based storage as in Freenet.
- southerntofu 5y ago
  Yes P2P CAS (torrent/freenet/ipfs) sounds very interesting over mixnets. Do you have links to specific implementations?
  - summm 5y ago
    Freenet manages to implement some kind of anonymity without mixnets, relying on inserting/retrieving storage blocks sacrificing latency. In my opinion that's the better approach, because if you separate storage and communication, like ipfs-over-tor, you'll never get anything with proper anonymity and decent performance. Beyond that, I've heard of xolotl related to gnunet, but I am not sure to what extent it is actually implemented and/or working, as usual with gnunet.
    - southerntofu 5y ago
      Are you aware of modern Freenet reimplementations, by any chance?
- goodpoint 5y ago
  content-based storage is not "instead" of TCP. It's a different solution for a different requirement.
  - southerntofu 5y ago
    correct, content-addressed storage is often implemented on top of TCP. I wrongly used TCP to generically mean "location-addressed".