Connecting Robust Shuffle Privacy and Pan-Privacy

Balcer, Victor; Cheu, Albert; Joseph, Matthew; Mao, Jieming

doi:10.48550/arxiv.2004.09481

Cited by 1 publication

(3 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In general, if each node over G p has at least one trustful neighbor, the privacy of the nodes will have further guarantee in terms of identifiability of β i (0), i ∈ V from β i (S), i ∈ V in the presence of malicious nodes. Therefore, the structure of G p would enable stronger internal privacy preservation that goes beyond Theorem 1, due to the shuffling effect [35,36] that comes along the PPSC procedure. We leave a quantitive analysis for this PPSC enabled internal privacy protection in future works since it is not fully aligning with the scope of the current paper.…”

Section: Discussion: Privacy Within the Private Graphmentioning

confidence: 99%

“…Recent advances on gossiping protocols include new privacy-preserving gossip algorithms [34]. On the other hand, the idea of shuffling data for differential privacy is also studied in [35,36], where each agent randomizes its own local data, and then submits the resulting randomized data to a secure shuffler for random permutation before being public for computation purpose. In such a protocol, a central shuffler is needed, and as in the previous differentially private computing protocols, it leads to a trade-off between the privacy and accuracy.…”

Section: Related Workmentioning

confidence: 99%

“…Note that in each iteration of the multi-gossiping PPSC mechanism, a pair of gossiping nodes essentially "shuffle" their states in a specific way: one node holds a noisy summation of the two node states, and another node holds a noise correlated to the first node's new state. This is related, but different from the idea of shuffling data [35,36], where randomized local data is sent to a central secure shuffler, and then the shuffler applies random permutations before being public for the computation purpose. The multi-gossiping PPSC mechanism does not rely on central secure shufflers; and the outcomes of the nodes states after the multi-gossiping PPSC mechanism and permutation shuffling are not the same.…”

Section: At Each G Kmentioning

confidence: 99%

See 2 more Smart Citations

Differentially Private Distributed Computation via Public-Private Communication Networks

Wang¹,

Liu²,

Manchester³

et al. 2021

Preprint

View full text Add to dashboard Cite

This paper studies the problem of multi-agent computation under the differential privacy requirement of the agents' local datasets against eavesdroppers having access to node-to-node communications.We first propose for the network equipped with public-private networks. The private network is sparse and not even necessarily connected, over which communications are encrypted and secure along with the intermediate node states; the public network is connected and may be dense, over which communications are allowed to be public. In this setting, we propose a multi-gossip Privacy-Preserving/Summation-Consistent (PPSC) mechanism over the private network, where at each step, randomly selected node pairs update their states in such a way that they are shuffled with random noise while maintaining summation consistency. It is shown that this mechanism can achieve any desired differential privacy level with any prescribed probability. Next, we embed this mechanism in distributed computing processes, and propose privacy-guarantee protocols for three basic computation tasks, where an adaptive mechanism adjusts the amount of noise injected in PPSC steps for privacy protection, and the number of regular computation steps for accuracy guarantee. For average consensus, we develop a PPSC-Gossip averaging consensus algorithm by utilizing the multi-gossip PPSC mechanism for privacy encryption before an averaging consensus algorithm over the public network for local computations. For network linear equations and distributed convex optimization, we develop two respective distributed computing protocols by following the PPSC-Gossip averaging consensus algorithm with an additional projection or gradient descent step within each step of computation. Given any privacy and accuracy requirements, it is shown that all three proposed protocols can compute their corresponding problems with the desired computation accuracy, while achieving the desired differential privacy. Numerical examples are used to illustrate the validity of the established theoretical results. In particular, our framework demonstrates clear improvements in terms of learning accuracy for classification problems compared to existing approaches under the same privacy budget.

show abstract

Section: Discussion: Privacy Within the Private Graphmentioning

confidence: 99%