Off-line Karma

Universityof

Nijmegen

Flavio D. Garciajoint work with

Jaap-Henk HoepmanNijmegen Institute for Computing and

Information ScienceRadboud University Nijmegen

Off-line KarmaOff-line Karma

A Decentralized Currency for P2P and Grid Applications

22/04/23Flavio Garcia - SOS Group - Off-line Karma2/21

OverviewOverview

Peer-to-peer systems Application example P2P Applications

Related work Our approach

For static networks For dynamic networks

Conclusions

22/04/23Flavio Garcia - SOS Group - Off-line Karma3/21

Peer-to-peer networksPeer-to-peer networks[Castro et al ’02] “Secure routing for structured Peer-to-Peer overlay networks”

Id, mCA

22/04/23Flavio Garcia - SOS Group - Off-line Karma4/21

f(x)x

Application ExampleApplication Example

f(x)xx

S

22/04/23Flavio Garcia - SOS Group - Off-line Karma5/21

P2P ApplicationsP2P Applications

Seti@home Folding@home Distributed.net Napster Gnutella Kazaa eDonkey

Free riding problem

Free riding problem

Donations

Donations

22/04/23Flavio Garcia - SOS Group - Off-line Karma6/21

Related WorkRelated Work

Generic Payment Schemes Rivest and Shamir’s “Pay World” Rivest’s “Electronic lottery” Rivest and Shamir’s “Micromint” Multi-payment schemes based on

public key signatures P2P Specific

The POPCORN project MojoNation KARMA PPay

22/04/23Flavio Garcia - SOS Group - Off-line Karma7/21

Objectives:Objectives: Scalability Transaction cost should be

independent of the size of the network. No centralized control The system should

not rely on one or several central, nodes and should not require any hierarchy.

Load Balance The overhead of the protocol is, in average, equally distributed over the peers.

Availability Two Karma users should be able to make transactions even when users are joining or leaving.

Double-spending detection For every double spent coin, a fraudulent user should be blacklisted

22/04/23Flavio Garcia - SOS Group - Off-line Karma8/21

The Threat ModelThe Threat Model

22/04/23Flavio Garcia - SOS Group - Off-line Karma9/21

The Threat ModelThe Threat Model

22/04/23Flavio Garcia - SOS Group - Off-line Karma10/21

Off-line Karma for Static Off-line Karma for Static NetworksNetworks

MintingSpendingRe-minting

22/04/23Flavio Garcia - SOS Group - Off-line Karma11/21

22/04/23Flavio Garcia - SOS Group - Off-line Karma12/21

0

1

2

4b4a

3a 3b

0

3b3a

2

1

4a 4b

?

SpendinSpendingg

22/04/23Flavio Garcia - SOS Group - Off-line Karma13/21

Re-mintingRe-minting

DB

Check for fraud in DB

Coin expires |R| = r Remint set for

k is always the same

22/04/23Flavio Garcia - SOS Group - Off-line Karma14/21

22/04/23Flavio Garcia - SOS Group - Off-line Karma15/21

Dynamic NetworksDynamic Networks

Minting still works. Spending still works. Reminting has a problem:

Some reminter may be off-line. Remint set depends on density.

While verifying users do not know the state of the network at re-mint time. Adversary has more freedom.

22/04/23Flavio Garcia - SOS Group - Off-line Karma16/21

Fixing re-minting for dynamic Fixing re-minting for dynamic networksnetworks

DB

22/04/23Flavio Garcia - SOS Group - Off-line Karma17/21

adv. pick r nodes

make c faulty

How large should r be?

22/04/23Flavio Garcia - SOS Group - Off-line Karma18/21

Early double-spending Early double-spending detectiondetection

22/04/23Flavio Garcia - SOS Group - Off-line Karma19/21

Conclusions - DisadvantagesConclusions - Disadvantages

Public key crypto involved. Users need to be on-line

periodically. Coin grow. Re-minting is an expensive

operation.

22/04/23Flavio Garcia - SOS Group - Off-line Karma20/21

Conclusions - AdvantagesConclusions - Advantages

We have developed a secure and fully distributed currency.

We provide a non-repudiable blacklisting mechanism.

The message complexity is twotwo for spending and O( sO( s )) for re-minting.

Possible early double-spending detection.

22/04/23Flavio Garcia - SOS Group - Off-line Karma21/21

THE ENDTHE END