We describe a scheme that secures the remote execution of sequential computations in grid-computing scenarios. To the best of our knowledge, this is the first contribution that addresses the security of generic sequential computations. By dividing sequential tasks into smaller subtasks and permuting them among participants, we show that our scheme facilitates the insertion of selective redundancy and/or pre-computed functions {ringers) that are indistinguishable from other computations. We analyze the security of this proposal and we demonstrate that our scheme enables the detection of individual and colluding malicious participants. In addition, we show that our scheme can be equally used to securely track the progress of remote execution. We further investigate the damages introduced by possible chaining of errors within the remote execution and we discuss recovery mechanisms to counter these challenges. We validate our findings both analytically and empirically via simulations.
展开▼
