-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 > > Only if you're ignoring the context within which the tool operates > > Well, I'm not sure which of us has the better grasp on reality.
What aspect of the user's context have I missed? > Have you ever actually spoken to a Chinese dissident about their needs? One thing I've found when doing requirements analysis is that end users are great for eliciting requirements, but often very poor at specifying exactly what they need. If it sounds good, they "need it". This is why requirements analysts are useful - they turn statements like "I need 4096bit RSA!!!" into requirements like "The user requires data confidentiality and integrity sufficient at preventing unwanted exposure to a state sized adversary for up to 50 years". It is then up to those providing a solution to meet those requirements to determine what the most appropriate technique is - maybe RSA4096 makes sense, but maybe not. "I NEED a globally scalable darknet" fits in that same category - while the user thought it sounds good, such a statement shouldn't be taken on face value. > > Its the same as arguing whether SCSI or IDE is better when > > the user is transferring data over a modem - its entirely academic. > I'm sorry, but its not, its the critical metric by which systems like > I2P and Freenet must be judged. If not, what criteria would you use > to measure the effectiveness of an anonymity-preserving system? A system is sufficiently anonymous if the cost of mounting an attack either exceeds the value to the attacker of success or the ability for the attacker to mount it. Anonymity is just a matter of simple economics. The system does not exist in a vaccum - the context in which it operates is critical to evaluating its effectiveness. The bottleneck in Freenet/dark's anonymity is in its ability to remain undetected, as if the users are detected, the darknet is no longer dark, at least not to those who can see what users are participating. The question is whether that bottleneck is larger or smaller than the anonymity offered by Freenet itself - all the HTL permutations, indirection, caching, etc. Additional techniques to improve the anonymity of the system beyond the limit imposed by the bottleneck is wasted (and, from an engineering perspective, counterproductive, since that likely means more code, which means more bugs). All techniques beyond that bottleneck are immaterial given the context which imposes the bottleneck, just as one's hard drive throughput is immaterial when you're limited to 7KBps. > Your argument seems to be: "If you create an effective anonymous > system, then everyone might use it, which would attract the unwanted > attention of a government, where if you only have crap anonymity > systems, then no one system will prevail, and the government will be > unable to find a single obvious target". Pretty much. I'd add on the clause "which would be able to pierce its anonymity" before the second comma. Oh, and I'd drop the "effective" part at the beginning, obviously (as thats what we're arguing about - whether it is effective). > This, if it really is what you are saying, is incredibly dumb, it is > pure security through obscurity. Effective security systems do not > get weaker just because an attacker can scrutinise how they work. You're right - it *is* security through obscurity. This is the difference between encryption, where security is offered by math and economics, and steganography, where security is offered by obscuring the participants. The difference between Freenet/dark and Freenet/light is that Freenet/dark is attempting to offer a steganographic network (which is different from a steganographic transport, which in turn is different from steganographic content). Theoretical question - would a large scale Freenet be useful in a hostile regime if the adversary knew who was participating, and when? If so, the "dark" is immaterial, and the system offers real value. If not, detecting who is participating in the network becomes one of two potential independent bottlenecks - 1) the anonymity and security offered by network's operation 2) the ability to remain dark (meaning undetected) If the cost of exposing (2) exceeds the cost of breaking (1), then work on (1) is worthwhile. However, if the cost of breaking (1) exceeds the cost of exposing (2), then work on improving (1) doesn't make sense, as the adversary will just go through the open window, ignoring the heavily fortified door. (2) is entirely security through obscurity - hoping that the adversary doesn't have the resources or inclination to mount an attack, not on the strength of their attack. I've yet to hear of any large scale steganographic technique capable of facing a plausible state level adversary. In fact, I've explained how the bottleneck exposed by (2) is relatively small to such an adversary - ISPs are already more than capable of offering the local monitoring, technical review, and administrative channels necessary to break it. Security through obscurity works, but only when no one is looking. Let me revise that - security through obscurity works, but only when no one willing and able to pierce the obscurity is looking. Small, "primitive" systems reduce the value of piercing the obscurity, which, given a resource constrained adversary (and they're all resource constrained), are less apt to draw the scrutiny that a larger "high tech" system would. As such, their (2) above would not be their bottleneck - instead, it would be their (1). > > One Big System which will not withstand focused attack is less secure > > than many small systems, even if each of those small systems would > > not be able to withstand the same level of attack. > If the resources required to attack the big system are far greater > than they would be to attack the individual small systems, as is the > case with a darknet as compared to the far less secure tools people > are using today (such as HTTP proxies, and I2P), then no that > argument certainly does not make sense. If the bottleneck of (2) is less constricting than the bottleneck of (1), I agree. However, as long as the bottleneck of (2) is less than the bottleneck of (1), the only meaningful analysis of the anonymity and security of the system is of (2). What makes you think the bottleneck of (1) is more constricting than the bottleneck of (2)? I've explained before why I believe the bottleneck exposed by Freenet/dark's (2) isn't much - is our disagreement based on that issue? Have I misjudged the ability of a state level adversary to pierce Freenet/dark's obscurity, or have I not yet made clear why I believe it isn't much? (as for the relative security of Freenet and I2P, lets save that for another day and another thread) > So, while from your perspective I can understand the attraction of an > argument that suggests that less effective anonymity tools might be > preferable to more effective anonymity tools, I doubt anyone without > a vested interest in a less effective anonymity tool would agree. I have no vested interest in an anonymity tool for use in hostile regimes. I2P is not appropriate for such users on a large scale. We're both reasonable people, and its seems we're both being sincere in expressing our thoughts - what fact or supposition do we differ upon to lead us so such divergent conclusions regarding the darknet's usefulness? =jr -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.1 (GNU/Linux) iD8DBQFDS80jWYfZ3rPnHH0RAkrcAJ92YxzErjUh4591TVrqhnE6yYXVzwCeIaFh yAjzhNtjTW0nL/juPTDkGTA= =h+yA -----END PGP SIGNATURE-----
