Re: CTP Issue 22: Authorization Token is obtained by truncating the results of the HMAC_SHA1 computation to retain only the leading 32 bits
Charlie Perkins <[email protected]>
| Newsgroups | gmane.ietf.seamoby |
|---|---|
| Organization | Nokia |
| Message-ID | <[email protected]> |
Hello Jim, Maybe it's time to insert rationality back into security protocol design. I'm willing to stand the heat. Mainly I am concerned about data fields that have to be passed over the air. Conservatism is indicated for this reason, and the reward is less hogging the airwaves. I am not worried about silicon, but in fact it would be nice if we did try to avoid mandating every protocol to require a Pentium processor. Regards, Charlie P. James Kempf wrote: >Hi Charlie, > >These are good arguments, but I don't see them swaying the security >directorate. Assuming one instruction per cycle and perhaps 100 instructions >to check, a 2 GHz Pentium can check about 20 million tokens per second. >Assuming 200 ms active time, that's about 4 million tokens in the time that >the token is active. > >One can argue with the details of the above calculation, but it is the kind >of thing I'd expect to see pop out of a security review. Of course, I >suppose we could argue that the document is Experimental anyway, so perfect >security isn't necessary, but what's the harm in lengthening the field? > > jak > >----- Original Message ----- >From: "Charlie Perkins" <[email protected]> >To: "James Kempf" <[email protected]> >Cc: <[email protected]> >Sent: Tuesday, September 16, 2003 12:34 PM >Subject: Re: [Seamoby] CTP Issue 22: Authorization Token is obtained by >truncating the results of the HMAC_SHA1 computation to retain only the >leading 32 bits > > > > >>Hello James, >> >>I don't think it is needed, for two reasons: >> >>- The authorization token is only going to be used once >> >>- The data being protected is only good for a few hundred >> milliseconds at most (typically). >> >>Regards, >>Charlie P. >> >> >>James Kempf wrote: >> >> >> >>>Issue 22 suggests that 32 bits isn't enough for the authorization token. >>> >>>See http://danforsberg.info:8080/draft-ietf-seamoby-ctp/issue22 for >>> >>> >details. > > >>>Possible resolutions: >>> >>>1) Make the authorization token longer, say 64 bits. >>>2) Keep the authorization token at 32 bits. >>> >>>Suggested resolution: Make the authorization token longer, say 64 bits. >>>Justification: The probability of a birthday attack is proportional to >>>2**(sqrt(n)), where n is the number of bits in the identifier. If n is >>> >>> >32, > > >>>then sqrt(n) is 16, which is not enough protection (65536). Making n >>> >>> >equal > > >>>64 means sqrt(n) is 32 (over 4 billion) which should be enough. >>> >>> >>>_______________________________________________ >>>Seamoby mailing list >>>[email protected] >>>https://www1.ietf.org/mailman/listinfo/seamoby >>> >>> >>> >>> >> >> >> > > >