Privacy-preserving techniques in recommender systems: state-of-the-art review and future research agenda
| DOI | https://doi.org/10.1108/DTA-02-2022-0083 |
| Published date | 04 May 2022 |
| Date | 04 May 2022 |
| Pages | 32-55 |
| Subject Matter | Library & information science,Librarianship/library management,Library technology,Information behaviour & retrieval,Metadata,Information & knowledge management,Information & communications technology,Internet |
| Author | Dhanya Pramod |
Privacy-preserving techniques
in recommender systems:
state-of-the-art review
and future research agenda
Dhanya Pramod
Symbiosis Centre for Information Technology,
Symbiosis International (Deemed University), Pune, India
Abstract
Purpose –This study explores privacy challenges in recommender systems (RSs) and how they have
leveraged privacy-preserving technology for risk mitigation. The study also elucidates the extent of adopting
privacy-preserving RSs and postulates the future direction of research in RS security.
Design/methodology/approach –The study gathered articles from well-known databases such as
SCOPUS, Web of Science and Google scholar. A systematic literature review using PRISMA was carried out on
the 41 papers that are shortlisted for study. Two research questions were framed to carry out the review.
Findings –It is evident from this study that privacy issues in the RS have been addressed with various
techniques. However, many more challenges are expected while leveraging technology advancements for fine-
tuning recommenders, and a research agenda has been devised by postulating future directions.
Originality/value –The study unveils a new comprehensive perspective regarding privacy preservation in
recommenders. There is no promising study found that gathers techniques used for privacy protection. The
study summarizes the research agenda, and it will be a good reference article for those who develop privacy-
preserving RSs.
Keywords Recommender, Privacy, Risk, Homomorphic cryptography, Differential privacy,
Privacy-preserving
Paper type Literature review
1. Introduction
Big data analytics has gained momentum in recent years, and many organizations are
leveraging it for profit-making. Recommender systems (RSs) are essential in item
recommendations to companies and customers, thereby enhancing business. RSs use
users’profiles, characteristics of items, interests of users and item usages for prediction and
recommendation, which are associated with users’privacy (Zhou et al., 2019). RS collects more
information from users to facilitate more accurate personalized recommendations. In recent
years, recommenders have been developed for various domains such as tourism, e-commerce,
entertainment, healthcare and social networking (Forouzandeh et al., 2015,2021,2022;
Aı€meur et al., 2008;Wang et al., 2019;Forouzandeh et al., 2021a,b). The RS helps users find
items and services that are of interest to them. In this context, hotel recommenders, movie
recommenders and product recommenders are evolved. The RS generally collects
demographic information such as age, gender, weight, educational background, behavioral
data such as activity status, location, browsing history, purchase pattern and rating history.
These are critical information and may lead to privacy risks if not handled properly, and the
risk goes very high if the data go to a third party without the user’s consent (Wang
et al., 2018).
Dealing with data is challenging in a RS, with privacy and security regulations. In this
context, there is a growing interest in designing privacy-preserving RSs (Hu et al., 2020).
Online business portals gather clients’personal information for various analytical insights,
and the data tampering attack on such sites is captivating. Such portals use RSs, and due to
DTA
57,1
32
Received 25 February 2022
Revised 4 April 2022
Accepted 18 April 2022
Data Technologies and
Applications
Vol. 57 No. 1, 2023
pp. 32-55
© Emerald Publishing Limited
2514-9288
DOI 10.1108/DTA-02-2022-0083
The current issue and full text archive of this journal is available on Emerald Insight at:
https://www.emerald.com/insight/2514-9288.htm
the lack of security controls, it becomes a vulnerable source of client information leakage.
User profiling leads to privacy threats. General Data Protection Regulation (GDPR) is a
regulation that came in to prevent the misuse of private information. Tejeda-Lorento et al.
(2018) suggested a set of guidelines for assessing and implementing GDPR-compliant RSs.
User-object interaction times may help understand user interest for a particular product,
and a system that detects such interactions is helpful for retail stores. In this case, context-
awareness and user-centric privacy preservation-based recommender are two essential
aspects to be integrated into such a system for better acceptance from the retail shops
(Parada et al., 2016). There have been attacks to leak customers’ratings, thereby violating
privacy.
Privacy challenges are therefore defined as “undesirable disclosure of users, their personal
preferences, data pertaining to users and selling users data without consent.”
With the popularity of personalized RSs, privacy concerns are multifold. Various
techniques to preserve privacy such as differential privacy, k-anonymity and privacy-
aware recommendation are used by many researchers to deal with security (Calero Valdez
and Ziefle, 2019). Personalized search systems create user profiles based on user interests
and collect user data. This personalized browsing experience raises serious security
concerns, and there is a dire need to devise a privacy-preserving system (El-Ansari et al.,
2022). It was also observedthat memory-based collaborative filtering RSs face many data
access challenges concerning privacy (Abid et al., 2020).
1.1 Related work
Many studies across domains deal with the privacy issues in RSs. Multiagent system
technology uses filtering techniques to make privacy-aware recommendations (Cissee and
Albayrak, 2007). Entropy-based randomness determination procedure could help in privacy
preservation in multicriteria collaborative filtering systems (Yargic and Bilge, 2019). The
weighted nonnegative matrix tri-factorization technique can differentiate genuine and fake
ratings. In this method, the trustworthiness of the customer ratings is assessed and thus
prevents privacy compromise (Wang et al., 2016).
PGuide is a pre-clinical guidance scheme that uses privacy-preserving comparison
protocol (PPCP) for hospital recommendation services (Wang et al., 2019). The computational
cost was found acceptable, and communication overhead was negligible. A harm-aware RS
allows users to make privacy-preserving decisions when using social media. Disclosure
appetite, a psychometric value and sensitivity of the revealed data are used to make privacy-
aware recommendations (Salem et al., 2021).
A privacy-protected RS (PRIPRO) was proposed for context-aware physical activity
recommendation (Sengan et al., 2021). A chaotic Reversible Data Transform (RDT)-based
algorithm for privacy-preserving data mining was proposed for PP in RS (Beg et al., 2021).
Tabassum and Ahmad (2021) proposed an iterative framework, with privacy preservation
in the reciprocal recommendation. The framework can be used for symmetric applications
like matrimonial, and asymmetric applications like job recommendations. Forouzandeh et al.
(2021a,b) emphasize that friendship-based user ratings can help in making recommendations
with limited data, and this trust-based rating helps in preserving privacy to an extent.
Blockchain-based RSs are becoming popular as they provide resilience, fault tolerance and
trust and are thus considered secure and p rivacy-enhancing (Himeur et al.,2022).
Advancements in the Internet of Things (IoT), digital transformation and artificial
intelligence (AI) have led to the evolution of RSs. RSs for energy efficiency in buildings
have gained attention in recent years and consume lots of data, including social and personal
(Himeur et al., 2021). YANA (You Are Not Alone) has been proposed for content
recommendation in online social communities (Li et al., 2011a). The primary aim here was
Privacy
techniques in
recommender
systems
33
To continue reading
Request your trial