wiki:NewCaps/Rainhill

Version 5 (modified by zooko, at 2012-06-20T14:41:59Z) (diff)

change self-link

This is about What Could Go Wrong with the "Rainhill 3" immutable file caps: http://jacaranda.org/tahoe/immutable-rainhill-3.svg

(http://jacaranda.org/tahoe/immutable-rainhill-3.png if your browser does not correctly handle SVG.)

http://jacaranda.org/tahoe/immutable-rainhill-3.png

The following table is based on a previous version of the protocol (Elk Point 2). It has been updated, but doesn't yet take into account attacks against the new features (incremental verification and deep-verify caps).

#what bad thing could happenhowwho could do itwhat could they targetwhat crypto property prevents ithow expensive to brute force [footnote 6]
1shape-shifter immutable file [footnote 1]collide read-cap (R,T)creator of a filetheir own filethe hash function's and cap format's collision resistance on the read-cap (R,T). This also depends on the hashes and PRP being correctly implemented, and on the suitability of hash_k as a KDF (key derivation function).approx sqrt(2.p).2(r+t)/2 [footnotes 4,5]
2unauthorized readattack the encryption of K_R with Ranyoneany one filethe security of the PRP, the secrecy of the read-key R, and the suitability of hash_r as a KDF.p.2min(r,k) / m
3forgery of immutable filegenerate a matching read-cap (R,T) for someone else's fileanyoneany one filethe hash function's and cap format's second-preimage resistance on (R,T). This also depends on the hashes and PRP being correctly implemented, and on the suitability of hash_k and hash_r as a KDF.p.2r+t / m [footnotes 3,5]
4unauthorized readattack the encryption of the plaintext with K_Ranyoneany one filethe security of the encryption scheme used for the plaintext, and the secrecy of the encryption key K_R. The latter also depends on the security and seeding of the RNG that generated it, and on resistance to attack #2.p.2k / m
5unauthorized readfigure out the input to the hash function that generates S and/or Vanyoneany one filethe hash function's onewayness for R -> V or V -> Sbrute force on R is #2
6accidental collisionstorage indices (S1,T1) and (S2,T2) collide accidentallynot applicableany two filesapproximately random distribution of hash function outputs[footnote 2]
7denial of serviceprevent access to servers holding sufficient shares (by controlling some of them, or by attacking them or the network)anyoneany filenot prevented by cryptonot applicable
8cause invalid share to verifygenerate (EncK_R,EncK_V,Q) that hashes to someone else's T, and copy their Sanyoneany one filethe hash function's second-preimage resistance on Tp.2t / m [footnote 3]

where k = bitlength(K_*), r = bitlength(R), s = bitlength(S) = bitlength(V), t = bitlength(T).

p is the success probability of an attack (0 < p <= 1). m is the number of targets for preimage attacks; this assumes that the attacker has stored the relevant hashes for m files and is content with finding a preimage for any of them. Note that since the attacker must also expend work to obtain each target hash, the cost of brute force cannot be reduced below a "square root" attack. For example, the work to forge an immutable file (attack #3) by brute force cannot be reduced to less than sqrt(p).2(r+t)/2 (roughly the same work as a collision attack), no matter how many targets are available.

  1. shape-shifter immutable file: creator creates more than one file matching the immutable file readcap
  1. See the probability table at http://en.wikipedia.org/wiki/Birthday_Attack . The effective hash length is approximately r+t bits.
  1. On Merkle-Damgård hashes with an internal state that is the same size as the hash output (like SHA-256), there are better second-preimage attacks than brute force. See http://www.schneier.com/paper-preimages.pdf . The doubled "SHA-256d" construction used by Tahoe does not help here. This is not significant for roadblock/speedbump attacks because the internal state will be much larger than t bits, but it is significant for the other second-preimage attacks.
  1. The formula given in the Wikipedia Birthday Attack page is sqrt(2.ln(1/(1-p))).2(r+t)/2, but the approximation given here is very accurate for small p, and can only underestimate the cost. For p = 1/2 it underestimates by only a factor of 1.18. For p near 1 it underestimates severely; it is very hard for an attacker to be certain to find a collision.
  1. In order for the combined hash with output (R,T) to have the strength against collision and preimage attacks given here, there must not be multicollision attacks against the hash truncated to r bits or to t bits, that would yield an easier attack on the combined hash. See tahoe-dev/2009-October/003006.html.
  1. The estimates given here are in terms of work factor, i.e. they are products of machine size and attack time. See this paper by Dan Bernstein for discussion of parallel brute-force attacks, including attacks against multiple keys at once.

Attachments (2)

Download all attachments as: .zip