Secure Distributed Data Aggregation
By Haowen Chan, Carnegie Mellon University, USA, haowenchan@cmu.edu | Hsu-Chun Hsiao, CyLab/Carnegie Mellon University, USA, hchsiao@cmu.edu | Adrian Perrig, CyLab/Carnegie Mellon University, USA, perrig@cmu.edu | Dawn Song, University of California, USA, dawnsong@cs.berkeley.edu
Abstract
We present a survey of the various families of approaches to secure aggregation in distributed networks such as sensor networks. In our survey, we focus on the important algorithmic features of each approach, and provide an overview of a family of secure aggregation protocols which use resilient distributed estimation to retrieve an approximate query result that is guaranteed to be resistant against malicious tampering; we then cover a second family, the commitmentbased techniques, in which the query result is exact but the chances of detecting malicious computation tampering is probabilistic. Finally, we describe a hash-tree based approach that can both give an exact query result and is fully resistant against malicious computation tampering.
Secure Distributed Data Aggregation
SSecure Distributed Data Aggregation surveys the various families of approaches to secure aggregation in distributed networks such as sensor networks. It focuses on the important algorithmic features of each approach, and provides an overview of a family of secure aggregation protocols which use resilient distributed estimation to retrieve an approximate query result that is guaranteed to be resistant against malicious tampering. It then covers a second family, the commitment-based techniques, in which the query result is exact but the chance of detecting malicious computation tampering is probabilistic. Finally, it describes a hash-tree based approach that can both give an exact query result and is fully resistant against malicious computation tampering.
In its selection of covered literature, Secure Distributed Data Aggregation sets out to provide the reader with a general intuitive understanding of the field, rather than to bring the reader exhaustively up to date with all algorithms for the area. It adopts a tutorial approach, selecting the publications that most clearly exemplify a certain class of approaches (or which have been most influential historically), rather than focusing on breadth or depth of coverage in terms of the most effective or the most recent algorithms.
