All the encryption layers for SAFEnet

[4M8B]

I think this is somewhat what you are seeing … currently we are only connected to one droplet ip this test.
Previous test was parallel to each other…

Consensus on SAFE

In tests only, you will have parallel connections at launch though. We do’t trust individual nodes so you will confirm via several (prb 3) routes. So it will mean more traffic, but probably necessary. Also 3 means we can cycle as these churn and clients should not be affected. Unless you connect through your own local vault So room for a vault to be integrated into launcher for ease and safety. it has connections to it’s group (plus normal kad routing table entries). Your messa…

[marmalade]

We should try and build a series of info graphics for this kinda stuff I think. One that explains the network, one that explains SAFEcoin, etc.

[feelz]

@Pierce

Plausible deniability also comes from the fact that chunks could just be flowing into you vaults without an explicit request by you. Just natural network healing as it churns.

Freenet claimed the same thing and a user got sybild. Though the salted part sounds like it may defeat chunks being id’d.

@Pierce

n addition, someone actually participating the network with the hopes of surrounding your connection in the network will need to have about 3 times the amount of vaults on the network to even hope to sybil your node.

With tor a hiddn service guard node could be knocked down and churn initiated to select attackers guard if not pinned and trusted, sybils are tricky business but if there is truly xor randomness and measures put in place to prevent attackers from spawning thousands of nodes in order to increase this probability (i believe i was part of this discussion and no one could come to a solution) then perhaps there is a chance maidsafe will be more anonymous and secure than tor.

[feelz]

Id like to see the mechanisms that prevent an attacker with x% of the nodes under its control from being selected as a guard as well as the rest of the chain. Just saying “XOR” doesnt explain anything, Tor implements pseudo-random node selection too (some basic checks like uptime, bandwidth, attack-scanning), the fact is without some way of verifying the legitimacy of a node, and with automated node-selection, you will always have x% chance of selecting a compromised chain of nodes.

Breaking up the requests so that its not a direct A->B->C->D->E chain like Tor has, and including default network routing (basically freenet), certainly does provide plausible deniability, but as you pointed out, that doesnt keep the cops from coming to your door and searching your house. They broke the denability by surrounding his node and monitoring known hashes (the packet fragments thereof) going into and out of his computer to determine if he was requesting the entirety of the file or just serving parts for others, turns out they were able to observe the former.

So salting in transit, if it does work in a truly random way, may in fact defeat this particular attack vector of a sybil, but it doesnt defeat sybil itself.

[anon40790172]

@feelz

Id like to see the mechanisms that prevent an attacker with x% of the nodes under its control from being selected as a guard as well as the rest of the chain.

On SAFE all nodes are guards. Even the slow ones that can’t route data chunks. Each nodes should at least be able to take part in a group and do simple things like signs and verifying signs from others. That way an evil node will be spotted quite fast. If you’re part of a group of 12 nodes, you need at least to own 7 or 8 of them to all do evil things. And you can’t pick your own address on SAFE. To target a group is extremely hard. Image you’re on IP-level and I’m in group XYZ and you want to target my group. How would you do that? You would ask the network to be accepted at XOR-level and as there are thousands of groups the change is extremely low you could join mine.

[goindeep]

Stupid question. Does the network use dynamic encryption keys?