TOR in anonimnost

zee je 29. jan 2012 ob 16:01 izjavil:

Ker "ubije" internetno hitrost.

- Slow. Especially when browsing .onion sites.
- TCP-based communication between nodes means more overhead, and the inability to distribute streams easily over nodes.
- Content to the internet can be spied on by exit nodes as well as modified while in transit. End-to-end encryption (such as SSL) is required. Don't see any real way to get around this. Besides, even your local ISP can inject data into your bitstream when browsing http sites "normally." Maybe all sites should be on SSL? But that would cause the "consumption" of IPv4 address space to increase rapidly. But this is another topic for debate.
- Heavy centralization. As of March, 2011, the entire Tor network is controlled by as little as 9 authorized directory servers that have unconditional rights to edit and enforce on regular participants lists of hidden services and lists of relay nodes. This raises serious questions related to anonymity and trust within the network. If and when those server operators (whose identities are well known - 9 people located in the USA and 4 European countries) decide to censor the Tor network (either by their own will, or by being forced by legislation or by arbitration or unlawfully), the entire network becomes vulnerable, can be easily censored, traced, or shut down. Regular members and even relay operators have no technical possibility to question or challenge authority of master directory operators, and despite representing about 99.65% of the Tor network users, are controlled by negligently small percentage of 0.35% of users. This is in sharp contrast to other Darknet projects, such as Freenet or I2P, where even overwhelming majority of participants have no technical possibilities to enforce their decisions even on negligently small minorities of users, and where there is no central authority to endanger.
- Not robust enough to carry torrent protocols and other traffic-heavy file transfers.

A more potent attack is a timing attack: If I watch Bob sending a request for a file, and then observe Alice getting a request of equal size, followed by Alice sending a 300MB file, if Bob gets a 300MB file, there is a good chance it might be Bob talking to Alice. This could be defeated with padding (making all data distributed on the network use a certain amount of data all the time), but that would be impractical and severely impact Tor's speed. This attack would be very useful in discovering the location of a Hidden Service, but it would take a very large amount of resources to successfully complete.