Scott Young: > Pinging the data would be as security-vunerable as requesting the > data itself.
NO NO NO! It is _much_worse_ because it circumvents our ideal bargain: "you can learn about the data, but I'll mirror it to 7 nodes in the process." I can gather all sorts of statistics by sending pings in various ways. > Also a malicious attacker would not be able to tell if a node had > certain data with this feature, because requests with htl=1 still > propate to other nodes but with certain probability (This is > already in the out-of-date documentation). This probability > should probably be increased if a node has a refrence to the file > in its datastore, but not the file itself. You're correct that this is our defense against htl=1 probes. I don't think it's implemented yet. > The security risk of multiple-pinging could be alleviated by > having nodes immediately return success if it already has a > refrence to the file in its datastore. This makes your ping even less useful, and it doesn't really help all that much - I'd be thrilled to get a list of nodes bearing references to my target data. The fact is, someone at the NSA is giggling like a schoolgirl whenever you suggest these things. _______________________________________________ freenet-tech mailing list [EMAIL PROTECTED] http://lists.freenetproject.org/mailman/listinfo/tech
