The redaction and authfailure-report documents are now in their second IETF Last Calls. So far, so good.
We can expect a comment from the IESG wondering why in the redaction document we didn't specify a particular hash algorithm. My reply so far (i.e., informally) is that it's not necessary. The agent generating the reports can select whatever hash it wants to use; if it's willing to risk collisions at the cost of cheaper processing, it can pick the weaker hashes. If it's satisfied with ROT13, it could even use that. The point here is to obscure the original string to the satisfaction of the report generator while allowing the report receiver to observe that multiple reports are referring to the same end user. The report receiver can then apply whatever tricks it wants to use to track the report back to the offending user once it gets a collection of such reports. Basically, the usual concerns about a collision attack don't apply to this use of hashes since the same party that produces the hashes also consumes them. So two questions: 1) Is that a reasonable reply? 2) Should the above be added as an Appendix? The AD I spoke to seems happy with this, and suggests that adding such text would help but it's not strictly necessary. -MSK
_______________________________________________ marf mailing list [email protected] https://www.ietf.org/mailman/listinfo/marf
