Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Decentralized Federated Learning: A Survey on Security and Privacy

Ehsan Hallaji, Roozbeh Razavi-Far, Mehrdad Saif, Boyu Wang, Qiang Yang

TL;DR

This survey studies possible variations of threats and adversaries in decentralized federated learning and overviews the potential defense mechanisms and Trustability and verifiability of decentralized federated learning are also considered.

Abstract

Federated learning has been rapidly evolving and gaining popularity in recent years due to its privacy-preserving features, among other advantages. Nevertheless, the exchange of model updates and gradients in this architecture provides new attack surfaces for malicious users of the network which may jeopardize the model performance and user and data privacy. For this reason, one of the main motivations for decentralized federated learning is to eliminate server-related threats by removing the server from the network and compensating for it through technologies such as blockchain. However, this advantage comes at the cost of challenging the system with new privacy threats. Thus, performing a thorough security analysis in this new paradigm is necessary. This survey studies possible variations of threats and adversaries in decentralized federated learning and overviews the potential defense mechanisms. Trustability and verifiability of decentralized federated learning are also considered in this study.

Decentralized Federated Learning: A Survey on Security and Privacy

TL;DR

This survey studies possible variations of threats and adversaries in decentralized federated learning and overviews the potential defense mechanisms and Trustability and verifiability of decentralized federated learning are also considered.

Abstract

Federated learning has been rapidly evolving and gaining popularity in recent years due to its privacy-preserving features, among other advantages. Nevertheless, the exchange of model updates and gradients in this architecture provides new attack surfaces for malicious users of the network which may jeopardize the model performance and user and data privacy. For this reason, one of the main motivations for decentralized federated learning is to eliminate server-related threats by removing the server from the network and compensating for it through technologies such as blockchain. However, this advantage comes at the cost of challenging the system with new privacy threats. Thus, performing a thorough security analysis in this new paradigm is necessary. This survey studies possible variations of threats and adversaries in decentralized federated learning and overviews the potential defense mechanisms. Trustability and verifiability of decentralized federated learning are also considered in this study.
Paper Structure (43 sections, 4 figures, 5 tables)

This paper contains 43 sections, 4 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Federated Learning
  4. Blockchain
  5. Decentralized Federated Learning
  6. Threats to Decentralized Federated Learning
  7. Attacks on Performance
  8. Data Poisoning
  9. Model Poisoning
  10. Routing Attacks
  11. Consensus Attacks
  12. Attacks on Privacy
  13. Model Inversion Attacks
  14. Membership Inference Attacks
  15. Hijacking Private Key
  16. ...and 28 more sections

Figures (4)

  • Figure 1: Workflow of distributed training, FL, and DFL with three nodes. $D_i$, $M_i$, and $u$ denote client data, local model parameters, and the generated update by the server. In this example, three clients are shown in the picture (i.e., $1\leq i\leq 3$).
  • Figure 2: Generic process of blockchain-based DFL. SC can use any choice of aggregation mechanism such as selecting the update with the highest score or averaging all received model parameters. $S_i$ indicates the calculated score for the estimated model parameters for client $i$.
  • Figure 3: Advances in decentralized federated learning in time. The order of the methods is set based on the first version that became available online (e.g., pre-prints).
  • Figure 4: Overview of security and privacy threats on decentralized federated learning. Threats are generally related to privacy, model robustness, or blockchain bottleneck of DFL.