Dear Andrew, I have created an extended version of my collision detection library [1] that, instead of a small selection of best disturbance vectors for SHA-1 collision attacks, simply tests all disturbance vectors within the two classes (see p125 of [2]).
This should rule out *all* cryptanalytic collision attacks that can be build from *any* disturbance vector mentioned in the literature (and more). I have run it over the initial set of 8 certs given by TSYS and the latest set of 7 certs, no collision attacks were detected. Note that these results are only valid IF the hash-then-sign SHA1-RSA signatures for these certs use the exact SHA-1 hashes below. I have not checked whether there is any further X.509/ASN1 encapsulation of the ToBeSignedPart in these files that should have been removed before checking. [stevens TSYS1]$ time ../detectcoll_allDVs *.tbs sha1 6ead26663275c388662dfdbc23ff0a76cdcf74dc ssl1.tsysacquiring.net.1.tbs sha1 3365793f36c197047b2f595c0f85c67b807c765f ssl1.tsysacquiring.net.2.tbs sha1 3c47155a5d9880a6893925e1c4479f914b3b9ffe ssl1.vitalps.net.1.tbs sha1 d130d1a8c51bce7323ba984b2f6298d0750405f4 ssl1.vitalps.net.2.tbs sha1 c9eb52c30bfaaaba5a1e060723e3c9e4b25f7b1e ssl2.vitalps.net.1.tbs sha1 3698794f1cabc3036380cc2adbc2805393098c45 ssl2.vitalps.net.2.tbs sha1 e7233e69a89b6b7568f790482b73f635d2464a95 ssl3.vitalps.net.1.tbs sha1 9c7bbae0fc9b08c5304908bd956c5b4b37c4e1c7 ssl3.vitalps.net.2.tbs Checked 5120 disturbance vectors real 0m0.134s user 0m0.132s sys 0m0.002s [stevens TSYS2]$ time ../detectcoll_allDVs *.der sha1 f75716390925b752b403a7bbf6acb349de9d8d09 ssl1.tsys.1.txt.der sha1 683c27025922c33c5dffb0c378368f09cd68acab ssl1.tsys.2.txt.der sha1 6740101673f95f50d97da18ce5d5d5d6e9bfe4b0 ssl1.vitalps.1.txt.der sha1 a4e0b6159b89f505d12b786a5562e808a7857786 ssl1.vitalps.2.txt.der sha1 24df6cf6d89a33f5262924169ff1a26ae69d49e8 ssl2.vitalps.1.txt.der sha1 57f51b7558805a953505426ee6757606bc13609d ssl2.vitalps.2.txt.der sha1 6d56564822a633f643f73b812ef0f86637af047d ssl3.vitalps.1.txt.der Checked 5120 disturbance vectors real 0m0.137s user 0m0.137s sys 0m0.000s Note that despite one fewer cert processing the second set takes a bit longer because the overall length is longer: [stevens TSYS1]$ cat *.tbs | wc -c 8468 [stevens TSYS2]$ cat *.der | wc -c 9334 Best regards, Marc Stevens [1] see https://marc-stevens.nl/research/#software [1] https://marc-stevens.nl/research/papers/phdthesis.pdf On 2016-07-19 22:04, Andrew R. Whalley wrote: > Greetings, > > I have run the tool provided by dr.ir <http://dr.ir>. Marc Stevens [1] on the > tbsCertificates provided by Symantec [2] > > And see no evidence of collisions: > > $ ./sha1dcsum_partialcoll *.tbs > 6ead26663275c388662dfdbc23ff0a76cdcf74dc ssl1.tsysacquiring.net.1.tbs > 3365793f36c197047b2f595c0f85c67b807c765f ssl1.tsysacquiring.net.2.tbs > 3c47155a5d9880a6893925e1c4479f914b3b9ffe ssl1.vitalps.net.1.tbs > d130d1a8c51bce7323ba984b2f6298d0750405f4 ssl1.vitalps.net.2.tbs > c9eb52c30bfaaaba5a1e060723e3c9e4b25f7b1e ssl2.vitalps.net.1.tbs > 3698794f1cabc3036380cc2adbc2805393098c45 ssl2.vitalps.net.2.tbs > e7233e69a89b6b7568f790482b73f635d2464a95 ssl3.vitalps.net.1.tbs > 9c7bbae0fc9b08c5304908bd956c5b4b37c4e1c7 ssl3.vitalps.net.2.tbs > > I'd be interested to know if anybody else replicates this. > > Marc - I believe that the tool as posted doesn't give assurance to the full > 80 bit security level. If that's true do you have an estimate of the > security level it does provide? > > Many thanks, > > Andrew > > [1] see https://marc-stevens.nl/research/ & > https://svn.marc-stevens.nl/collisiondetection/ > [2] https://cabforum.org/pipermail/public/2016-July/007999.html >
signature.asc
Description: OpenPGP digital signature
_______________________________________________ dev-security-policy mailing list dev-security-policy@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-security-policy
