Mitigating the ICA Attack against Rotation-Based Transformation for Privacy Preserving Clustering

Mohaisen, Aziz; Hong, Dowon

doi:10.4218/etrij.08.0208.0134

Cited by 6 publications

(18 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As an elementary representation, a 2×2 orthonormal matrix can be represented as R=f(θ)=[e ij ], where e ij = cosθ iff i=j, e ij = sinθ iff i<j, and e ij = -sinθ iff i>j. An aⅹa orthonormal matrix, where a is an even number, is expressed in terms of a 2×2 orthonormal matrix as a diagonal block matrix R = [R i ] where 1≤ i ≤ a/2 and R 1 = R 2 = … = R a/2 [4]. For odd numbers of attributes, the last attribute can be transformed manually with any other previously transformed attribute [5].…”

Section: A Rbtmentioning

confidence: 99%

“…In spite of mitigating the AK-ICA attack, which is a great merit [4], the MRBT does not allow conventional clustering on data that belongs to two different subsets. To enable that, we introduce an ARBT for conventional clustering of transformed data and its privacy.…”

Section: B Mrbtmentioning

confidence: 99%

“…On the other hand, the third phase is performed when clustering results are to be computed over two data subsets using the conventional clustering method. In the description below, we consider a single user model though it should be clear that extension to a multi-user model is straightforward as shown in [4]. The three different phases of the client side are detailed in Fig.…”

Section: Client Sidementioning

confidence: 99%

“…The data miner follows the same procedure as in MRBT in response to steps 1 and 2 of the client side: the server will be able to learn clusters over each data subset separately [4]. However, after the client performs step 3 and releases the parameter θ ij upon the server's request, the server performs the following.…”

Section: Server Sidementioning

confidence: 99%

“…To mitigate this attack, Mohaisen and others introduced a multiple rotationbased transformation (MRBT) algorithm [4]. Though the MRBT mitigates the AK-ICA attack and allows several distancebased algorithms including special kinds of clustering, it does not enable the conventional clustering on the transformed data [4].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Augmented Rotation-Based Transformation for Privacy-Preserving Data Clustering

Hong¹,

Mohaisen²

2010

ETRI J

Self Cite

View full text Add to dashboard Cite

Multiple rotation-based transformation (MRBT) was introduced recently for mitigating the apriori-knowledge independent component analysis (AK-ICA) attack on rotation-based transformation (RBT), which is used for privacy-preserving data clustering. MRBT is shown to mitigate the AK-ICA attack but at the expense of data utility by not enabling conventional clustering. In this paper, we extend the MRBT scheme and introduce an augmented rotation-based transformation (ARBT) scheme that utilizes linearity of transformation and that both mitigates the AK-ICA attack and enables conventional clustering on data subsets transformed using the MRBT. In order to demonstrate the computational feasibility aspect of ARBT along with RBT and MRBT, we develop a toolkit and use it to empirically compare the different schemes of privacy-preserving data clustering based on data transformation in terms of their overhead and privacy.

show abstract

Section: A Rbtmentioning

confidence: 99%

Section: B Mrbtmentioning

confidence: 99%

Section: Client Sidementioning

confidence: 99%

Section: Server Sidementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Augmented Rotation-Based Transformation for Privacy-Preserving Data Clustering

Hong¹,

Mohaisen²

2010

ETRI J

Self Cite

View full text Add to dashboard Cite

show abstract

A Computationally-Efficient Construction for the Matrix-Based Key Distribution in Sensor Network

Mohaisen

Jho

Hong

2009

Advances in Information Security and Assurance

Self Cite

View full text Add to dashboard Cite

This paper introduces a variant for the symmetric matrix-based key distribution in sensor network introduced by Du et al. Our slight modification shows that the usage of specific structures for the public matrix instead of fully random matrix with elements in Z q can reduce the computation overhead for generating the public key information and the key itself. An intensive analysis followed by modified scheme demonstrates the value of our contribution in relation with the current work and show the equivalence of the security G c (j)[j mod λ] ← g 1j .

show abstract

Aligned Space Time Block Codes for the 2-User X Channel with Secrecy Constraints

Mohaisen

2015

International Journal of Antennas and Propagation

View full text Add to dashboard Cite

Interference alignment (IA) is a technique used to reduce the dimension of the interference, where consequently the multiplexing rate is increased. In the 2-user X channel, combining IA with space-time block codes increases the diversity gain. These gains are achieved with the cost of leaked information at unintended receivers, where this leaked information can be used to decode other receiver’s signals. In this paper, we consider each of the two two-antenna receivers as an eavesdropper with 1 or 2 additional eavesdropping antennas. As such, we suggest receiver structures to answer the question: “Is the leaked information sufficient to properly decode the unintended signals?” besides quantifying the leaked information in terms of secrecy sum rates (SSR). Interestingly, we show that the SSR is negative, indicating that the quality of the eavesdropped signals is superior to that of the intended signals. To assure confidentiality, we propose an interleaved multiple rotation-based transformation scheme that neutralizes any a priori knowledge about the structure of the eavesdropped information and rotates the transmitted symbols using orthogonal matrices, preserving both the power and the distance between symbols.

show abstract

Mitigating the ICA Attack against Rotation-Based Transformation for Privacy Preserving Clustering

Cited by 6 publications

References 7 publications

Augmented Rotation-Based Transformation for Privacy-Preserving Data Clustering

Augmented Rotation-Based Transformation for Privacy-Preserving Data Clustering

A Computationally-Efficient Construction for the Matrix-Based Key Distribution in Sensor Network

Aligned Space Time Block Codes for the 2-User X Channel with Secrecy Constraints

Contact Info

Product

Resources

About