01-02-2013, 04:07 PM
Payments for Outsourced Computations
1Payments for Outsourced.docx (Size: 14.63 KB / Downloads: 21)
Abstract:
With the recent advent of cloud computing, the concept of outsourcing computations, initiated by volunteer computing efforts, is being revamped. While the two paradigms differ in several dimensions, they also share challenges, stemming from the lack of trust between outsourcers and workers. In this work, we propose a unifying trust framework, where correct participation is financially rewarded: neither participant is trusted, yet outsourced computations are efficiently verified and validly remunerated. We propose thee solutions for this problem, relying on an offline bank to generate and redeem payments; the bank is oblivious to interactions between outsourcers and workers. We propose several attacks that can be launched against our framework and study the effectiveness of our solutions. We implemented our most secure solution and our experiments show that it is efficient: the bank can perform hundreds of payment transactions per second and the overheads imposed on outsourcers and workers are negligible.
Existing System:
We consider a general “compute market” framework, encompassing the cloud and volunteer computing paradigms: participating computers can act both asservice providers (workers) and as clients (outsourcers).Outsourcers have computing jobs they cannot complete in atimely fashion, whereas workers are willing to spend CPU cycles to run parts of such jobs. While it is natural to
Motivate participation through the use of financial incentives; the distributed nature of the framework raises trust questions: Outsourcers do not trust the workers to correctly perform computations and workers do not trust outsourcersto pay for completed jobs. While solutions exist that address the lack of trust of outsourcers on workers, the lack of trust of a worker in the outsourcer is not addressed is required to fully trust. In settings where any cellphone or PC owner can be an outsourcer, this becomes a problem.
Proposed System:
We propose solutions that address both issues of trust. We rely on a trusted offline third party, a bank B, that acts strictly as a financial institution in the transaction between O and W. B issues payment tokens, which O embeds in jobs. W is able to retrieve a payment token if and only if it completes a job. Our solutions employ the ringer concept proposed by Golle and Mironov .Our first solution requires O to split the key used to obfuscate the payment and hides the sub keys into precompiled, randomly chosen parts of the job. The workeris entitled to a probabilistic verification of the payment received before beginning the computation.