Recommender systems (RS) are an integral part of many online services aiming to provide an enhanced user-oriented experience. Machine learning (ML) models are nowadays broadly adopted in modern state-of-the-art approaches to recommendation, which are typically trained to maximize a user-centred utility (e.g., user satisfaction) or a business-oriented one (e.g., profitability or sales increase). They work under the main assumption that users’ historical feedback can serve as proper ground-truth for model training and evaluation. However, driven by the success in the ML community, recent advances show that state-of-the-art recommendation approaches such as matrix factorization (MF) models or the ones based on deep neural networks can be vulnerable to adversarial perturbations applied on the input data. These adversarial samples can impede the ability for training high-quality MF models and can put the driven success of these approaches at high risk. As a result, there is a new paradigm of secure training for RS that takes into account the presence of adversarial samples into the recommendation process. We present adversarial machine learning in Recommender Systems (AML-RecSys), which concerns the study of effective ML techniques in RS to fight against an adversarial component. AML-RecSys has been proposed in two main fashions within the RS literature: (i) adversarial regularization, which attempts to combat against adversarial perturbation added to input data or model parameters of a RS and, (ii) generative adversarial network (GAN)-based models, which adopt a generative process to train powerful ML models. We discuss a theoretical framework to unify the two above models, which is performed via a minimax game between an adversarial component and a discriminator. Furthermore, we explore various examples illustrating the successful application of AML to solve various RS tasks. Finally, we present a global taxonomy/overview of the academic literature based on several identified dimensions, namely (i) research goals and challenges, (ii) application domains and (iii) technical overview.

Adversarial machine learning in recommender systems (AML-RECSYS) / Deldjoo, Yashar; Di Noia, Tommaso; Merra, Felice Antonio. - ELETTRONICO. - (2020), pp. 869-872. (Intervento presentato al convegno 13th ACM International Conference on Web Search and Data Mining, WSDM 2020 tenutosi a Houston, TX nel February 3-7, 2020) [10.1145/3336191.3371877].

Adversarial machine learning in recommender systems (AML-RECSYS)

Yashar Deldjoo;Tommaso Di Noia;Felice Antonio Merra
2020-01-01

Abstract

Recommender systems (RS) are an integral part of many online services aiming to provide an enhanced user-oriented experience. Machine learning (ML) models are nowadays broadly adopted in modern state-of-the-art approaches to recommendation, which are typically trained to maximize a user-centred utility (e.g., user satisfaction) or a business-oriented one (e.g., profitability or sales increase). They work under the main assumption that users’ historical feedback can serve as proper ground-truth for model training and evaluation. However, driven by the success in the ML community, recent advances show that state-of-the-art recommendation approaches such as matrix factorization (MF) models or the ones based on deep neural networks can be vulnerable to adversarial perturbations applied on the input data. These adversarial samples can impede the ability for training high-quality MF models and can put the driven success of these approaches at high risk. As a result, there is a new paradigm of secure training for RS that takes into account the presence of adversarial samples into the recommendation process. We present adversarial machine learning in Recommender Systems (AML-RecSys), which concerns the study of effective ML techniques in RS to fight against an adversarial component. AML-RecSys has been proposed in two main fashions within the RS literature: (i) adversarial regularization, which attempts to combat against adversarial perturbation added to input data or model parameters of a RS and, (ii) generative adversarial network (GAN)-based models, which adopt a generative process to train powerful ML models. We discuss a theoretical framework to unify the two above models, which is performed via a minimax game between an adversarial component and a discriminator. Furthermore, we explore various examples illustrating the successful application of AML to solve various RS tasks. Finally, we present a global taxonomy/overview of the academic literature based on several identified dimensions, namely (i) research goals and challenges, (ii) application domains and (iii) technical overview.
2020
13th ACM International Conference on Web Search and Data Mining, WSDM 2020
9781450368223
Adversarial machine learning in recommender systems (AML-RECSYS) / Deldjoo, Yashar; Di Noia, Tommaso; Merra, Felice Antonio. - ELETTRONICO. - (2020), pp. 869-872. (Intervento presentato al convegno 13th ACM International Conference on Web Search and Data Mining, WSDM 2020 tenutosi a Houston, TX nel February 3-7, 2020) [10.1145/3336191.3371877].
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11589/196493
Citazioni
  • Scopus 31
  • ???jsp.display-item.citation.isi??? 26
social impact