Preemptive Measures against Malicious Party in Privacy-Preserving Data Mining

Abstract

Currently, many privacy-preserving data mining (PPDM) algorithms assume the semi-honest model and/or malicious model of multi-party interaction. However, both models are far from able to fully capture the complexity of events and data interactions among parties in the process of secure data mining. In the paper, we study the problem of security violations when a malicious party provides false data. We identify four privacy vulnerabilities of secure scalar product protocols that underlie many current PPDM algorithms. We propose a general more model of two-party interaction and demonstrate its applicability to securely compute (x 1 + y 1)(x 2 + y 2) and (x + y) log 2 (x + y) where x i and y i are private values held by each party respectively. We show how the proposed model can be used to securely compute four commonly used kernel functions and other common functions. We also propose two necessary conditions and two basic measures that should be adopted in the current malicious model.

Incentive Compatible Privacy-Preserving Data Analysis

“…In addition to existing techniques that consider honestbut-curious model, there are techniques developed against malicious adversaries, such as [16], [24]. Especially, in [16], authors discuss how to prevent lying about inputs using "input-consistency checks."…”

“…Especially, in [16], authors discuss how to prevent lying about inputs using "input-consistency checks." Basically, authors suggest checking whether the inputs satisfy some conditions that are known to be true about the inputs (e.g., a binary input vector cannot consist of all zeros).…”

Distributed classification in P2P networks

“…(c) P I and P J securely and jointly compute δ = (g 1 + g 2 ) × (k 1 + k 2 ) = R 1 + R 2 using Random Shares Protocol [HN08] where R 1 and R 2 are held by P I and P J respectively. P I then sends R 1 to P J and P J sends R 2 to P I and they individually compute δ = R 1 + R 2 .…”

“…Linear kernel is used in this work which simplifies the scalar products, hence no other addition computation is necessary. However, note that other kernel functions can also be securely computed as shown by Han and Ng[HN08]. Here, the secure scalarproduct protocol as proposed by Goethals et al [GLLM05] is used.…”

Privacy-preserving data mining via secure multiparty computation

“…In the following, we discuss some issues which have not been explored in [53] regarding the use of Taylor series to approximate the inverse of scalar sum. To make the computation easier, if x 1/2 or y 1/2, we may choose sufficiently large values C so that 0 < x ′ = x/C < 1/2 and 0 < y ′ = y/C < 1/2 and that 0 <…”