This is probably still true. I2P has a lot more routing nodes than Tor does and a lot lot lot less clients than Tor does. It was also designed and optimized for hidden services. I would be surprised if Tor ever beats I2P for speed or scaleability. However Tor has made serious improvements to its hidden service load times since a few years ago (when this was written). This is kind of neat but many people would argue it makes it less secure than Tor. Tor uses centralized directory authority servers because they think it is required for anonymity. I2P people disagree. I am kind of on the fence regarding this issue, but I certainly see the problems with both systems. I am pretty sure Tor does this too now I2P people see this as a benefit, Tor people see this as a vulnerability. It isn't so black and white. security by obscurity Meh so is Tor if you know what you are doing. Do Tor circuits really fail so often? Dubious claim, research needs to be done to determine if I2P is more resistant to active correlation attacks than Tor is. And it isn't any safer from passive correlation attacks. If it is safer from active correlation attacks or not will have to do with every node being a router, not using unidirectional tunnels. It may very well actually double your risk of being pwnt by an end point timing attack, since if you connect to a malicious hidden service if they have active nodes at either your inbound or outbound tunnel they should be able to correlate packets to deanonymize you. Tor uses one node for inbound and outbound traffic. You only need a few high precision measurements to deanonymize someone with a timing attack if you can position yourself properly, I2P tunnels have two nodes to target (inbound and outbound) that each have half the traffic going through them than a Tor entry guard would....Tor has a single entry guard with twice the traffic on it. Unless someone can show me otherwise, I think Tor has the better design in this aspect. ] Tunnels in I2P are short lived, increasing the rate at which you will eventually use a node owned by X attacker. True How is this a benefit over Tor? I don't want to participate in routing for others. This is a benefit if you want a fast network that scales, not if you want anonymity as a client. Having all clients participate in routing for others is actually one of the biggest weaknesses of I2P as far as I am concerned, it makes it easier for a weaker attacker to get a list of all I2P peers IP addresses. At least with Tor as a client the attacker needs to monitor the directory authority servers, or your connection at your ISP, or your entry guard. I2P lets anyone who adds some high bandwidth nodes to the network to get everyone elses IP address. It also makes it so that if you have any down time your eepsite is as good as deanonymized. I also hope you are not in an IRC chat and running as an I2P router at the same time, because if you have down time you are going to be kicked from the IRC at the same time your I2P node goes down. Which is great if you want a fast network that scales but shitty if you want membership concealment as a client. It probably is also bad for client anonymity in other ways, running as a Tor relay actually decreases your anonymity if you use the same node as a client. Integrated automatic backdoor mechanism is not a benefit over Tor...and Torbutton lets you know when your Tor is out of date and lets you update it yourself This is certainly a benefit if you want to use UDP Yes Tor has many vulnerabilities and it sucks, but for anonymity it sucks a lot less than I2P does IMO. I would use Freenet before I used I2P. But I think Tor is the best choice. Also I2P is just as weak to all of those attacks as Tor is. Tor has had a fuck ton more research done on it than I2P has, but a lot of the attacks against Tor apply just as much to I2P.