Privacy-aware and scalable recommender systems uing sketching techniques

Abstract : In this thesis, we aim to study and evaluate the privacy and scalability properties of recommender systems using sketching techniques and propose scalable privacy preserving personalization mechanisms. Hence, the thesis is at the intersection of three different topics: recommender systems, differential privacy and sketching techniques. On the privacy aspects, we are interested in both new privacy preserving mechanisms and the evaluation of such mechanisms. We observe that the primary parameter in differential privacy is a control parameter and motivated to find techniques that can assess the privacy guarantees. We are also interested in proposing new mechanisms that are privacy preserving and get along well with the evaluation metrics. On the scalability aspects, we aim to solve the challenges arising in user modeling and item retrieval. User modeling with evolving data poses difficulties, to be addressed, in storage and adapting to new data. Also, addressing the retrieval aspects finds applications in various domains other than recommender systems. We evaluate the impact of our contributions through extensive experiments conducted on benchmark real datasets and through the results, we surmise that our contributions very well address the privacy and scalability challenges.
Document type :
Theses
Liste complète des métadonnées

Cited literature [135 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-01430156
Contributor : Abes Star <>
Submitted on : Monday, January 9, 2017 - 3:27:29 PM
Last modification on : Friday, January 11, 2019 - 3:15:16 PM
Document(s) archivé(s) le : Monday, April 10, 2017 - 3:07:48 PM

File

BALU_Raghavendran.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-01430156, version 1

Citation

Raghavendran Balu. Privacy-aware and scalable recommender systems uing sketching techniques. Other [cs.OH]. Université Rennes 1, 2016. English. ⟨NNT : 2016REN1S047⟩. ⟨tel-01430156⟩

Share

Metrics

Record views

309

Files downloads

419