Privacy in Deep Learning: A Survey

By Fatemehsadat Mirshghallah, Mohammadkazem Taram, Praneeth Vepakomma, Abhishek Singh, Ramesh Raskar, Hadi Esmaeilzadeh. 2020

153 citations recorded by the source corpus.

The ever-growing advances of deep learning in many areas including vision, recommendation systems, natural language processing, etc., have led to the adoption of Deep Neural Networks (DNNs) in production systems. The availability of large datasets and high computational power are the main contributors to these advances. The datasets are usually crowdsourced and may contain sensitive information. This poses serious privacy concerns as this data can be misused or leaked through various vulnerabilities. Even if the cloud provider and the communication link is trusted, there are still threats of inference attacks where an attacker could speculate properties of the data used for training, or find the underlying model architecture and parameters. In this survey, we review the privacy concerns brought by deep learning, and the mitigating techniques introduced to tackle these issues. We also show that there is a gap in the literature regarding test-time inference privacy, and propose possible future research directions.

Subfields: Artificial Intelligence, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Safety, Robustness & Interpretability

Citations in the corpus

Language Generation Models Can Cause Harm: So What Can We Do About It? An Actionable Survey (2022, background): “memorize these details and generate them verbatim when prompted by users, leading to privacy violations (Kim, 2016; Mirshghallah et al., 2020; Brown et al., 2022).”
FedML: A Research Library and Benchmark for Federated Machine Learning (2020, background): “While the FL framework facilitates data privacy [89] by keeping data locally available to the users and only requiring communication for model updates, users may still be concerned about partial leakage of their data which may be inferred from the communicated model (see, e.g. , [112]).”
Privacy Regularization: Joint Privacy-Utility Optimization in LanguageModels (2021, unknown): “This helps better distribute one user’s data in the embedding space. et al., 2017; Mirshghallah et al., 2020), and this slow training process makes the tuning of the parameters extremely harder.”
Addressing Privacy Threats from Machine Learning (2021, background): “A rapidly growing research area in computer science is that of privacy in machine learning (Papernot et al., 2018; Mirshghallah et al., 2020; Liu et al., 2021).”
Membership Inference Attacks on Machine Learning: A Survey (2021, background): “There are many surveys that summarize different attacks on ML models [35, 84, 114, 117, 137, 155, 166, 168, 176, 190, 228].”
Federated Learning for Open Banking (2021, background): “This chapter does not discuss details of privacypreserving techniques and data encryption as they have been well studied in many literature reviews [26,32].”
Differential Privacy for Deep and Federated Learning: A Survey (2022, background): “Fatemehsadat et al. [40] present a summary of information disclosure attacks to better situate the need for privacy protection in DL.”