Reconstructing Hash Reversal based Proof of Work Schemes

panelgameΑσφάλεια

3 Δεκ 2013 (πριν από 3 χρόνια και 10 μήνες)

95 εμφανίσεις

Joshua White
CS644
Professor: Dr. Jeanna Matthews
Short Paper Review of:
Reconstructing Hash Reversal based Proof of Work Schemes
Citation:
Jeff Green, Joshua Juen, Omid Fatemieh, Ravinder Shankesi, Dong Jin, and Carl A. Gunter. 2011.

Reconstructing hash reversal based proof of work schemes. In
Proceedings of the 4th USENIX

conference on Large-scale exploits and emergent threats
(LEET'11). USENIX Association, Berkeley,

CA, USA, 10-10.
Author(s) Affiliation(s):

University of Illinois at Urbana-Champaign

Funded by NSA, HHS, NASA, MacArthur Foundation, Boeing and Lockeed Martin.
Claims:

GPU resource-inflation attacks are a novel threat (not a claim but a generic statement).

An examination of work schemes based on hash reversal from a GPU-based attack standing.

Demonstrate hash reversal schemes that adapt soley on server load are ineffective under attack

by GPU utilizing adversaries.

Demonstrate that the hash reversal schemes that adapt based on client behavior are effective

under GPU based attacks.
Likes:

I found this paper to be particularly interesting because it has elements that apply to the bitcoin

system. While the paper its self doesn't mention bitcoin, the same issue exist in that ecosystem.

The basic idea behind the problem is simple, the server sends a puzzle to make sure that the

attacker can't send to many requests within a given period of time without solving a puzzle each

time. GPU's on the otherhand can do hashing calculations at much higher rates then CPU's

therefore the puzzling techniques that rely on how long the attacker takes to solve is ineffective.

Again the concept is simple but the subject its self is highly technical, I liked that point very

much.
Dislikes:

I wasn't able to find any associated website/source code for this paper/authors. I found that a bit

disappointing since I was interested in trying this one out on my own.

The authors chose to use an NVidia based GPU and Intel CPU for testing. I think this paper

would have been more comprehensive had they gone with an AMD CPU and ATI GPU. It's

well known at least in the bitcoin community that ATI GPU's are massivly more effective at

calculating hashes then any other GPU CPU in the normal consumer market.


https://en.bitcoin.it/wiki/Mining_hardware_comparison
” I'm actually surprised that they didn't

include a similar chart excerpt in the paper for discussion.
Repeat / Add to Work?

I thought this was a fairly comprehensive and technical paper on the subject. I think that the

authors make a good semi-cross discipline leap from DdoS and hash calculation systems which

are typically two disparate fields. However, I think they should have taken that cross-discipline

objective further and included another well known GPU hash calculation field (ie: bitcoin). I

think that theirs a lot of work in bitcoin security and hash calculation that could have applied to

this papers problem/objective.

While this was a very interesting paper to read and I found the topic applicable to other research

I've done, I don't see myself reproducing this paper or similar work anytime soon, it's a little to

hardware specific for my tastes. Anytime you can throw more hardware at a problem, my

interest fades.