FedSteg: A Federated Transfer Learning Framework for Secure Image Steganalysis

Yang, Hong-Wei; He, Hui; Zhang, Weizhe; Cao, Xiaochun

doi:10.1109/tnse.2020.2996612

Cited by 86 publications

(45 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…X j f eature , ∀D i , D j , i j. Yang et al [202] proposed a secure FTL framework, FedSteg, to train a personalized and distributed model for a secure image steganalysis. As its advantage, FedSteg is a general framework suitable for general network structures in privacy-preserving machine learning.…”

Section: Federated Transfermentioning

confidence: 99%

A Comprehensive Survey of Privacy-preserving Federated Learning

2021

View full text Add to dashboard Cite

The past four years have witnessed the rapid development of federated learning (FL). However, new privacy concerns have also emerged during the aggregation of the distributed intermediate results. The emerging privacy-preserving FL (PPFL) has been heralded as a solution to generic privacy-preserving machine learning. However, the challenge of protecting data privacy while maintaining the data utility through machine learning still remains. In this article, we present a comprehensive and systematic survey on the PPFL based on our proposed 5W-scenario-based taxonomy. We analyze the privacy leakage risks in the FL from five aspects, summarize existing methods, and identify future research directions.

show abstract

Section: Federated Transfermentioning

confidence: 99%

A Comprehensive Survey of Privacy-preserving Federated Learning

2021

View full text Add to dashboard Cite

show abstract

“…Fourthly, PFL via transfer learning, e.g., FedMD [12], FedSteg [13], and Fedhealth [14], is to either use model and domain-specific local fine-tuning or knowledge distillation to adapt the global model into the personalized model.…”

Section: Personalized Federated Learningmentioning

confidence: 99%

Achieving Personalized Federated Learning with Sparse Local Models

Huang¹,

Liu²,

Shen³

et al. 2022

Preprint

View full text Add to dashboard Cite

Federated learning (FL) is vulnerable to heterogeneously distributed data, since a common global model in FL may not adapt to the heterogeneous data distribution of each user. To counter this issue, personalized FL (PFL) was proposed to produce dedicated local models for each individual user. However, PFL is far from its maturity, because existing PFL solutions either demonstrate unsatisfactory generalization towards different model architectures or cost enormous extra computation and memory. In this work, we propose federated learning with personalized sparse mask (FedSpa), a novel PFL scheme that employs personalized sparse masks to customize sparse local models on the edge. Instead of training an intact (or dense) PFL model, FedSpa only maintains a fixed number of active parameters throughout training (aka sparse-to-sparse training), which enables users' models to achieve personalization with cheap communication, computation, and memory cost. We theoretically show that the iterates obtained by FedSpa converge to the local minimizer of the formulated SPFL problem at the rate of O( 1 √ T). Comprehensive experiments demonstrate that FedSpa significantly saves communication and computation costs, while simultaneously achieves higher model accuracy and faster convergence speed against several state-of-the-art PFL methods.

show abstract

“…As already said, S4T works independently from the applications deployed on it which can be written in many different programming languages. In [16], authors propose FedSteg, a FL framework for image steganalysis. Even if the obtained results are promising, the authors did not deploy it in a real scenario.…”

Section: Related Workmentioning

confidence: 99%

Leveraging Stack4Things for Federated Learning in Intelligent Cyber Physical Systems

Vita

Bruneo

2020

JSAN

View full text Add to dashboard Cite

During the last decade, the Internet of Things acted as catalyst for the big data phenomenon. As result, modern edge devices can access a huge amount of data that can be exploited to build useful services. In such a context, artificial intelligence has a key role to develop intelligent systems (e.g., intelligent cyber physical systems) that create a connecting bridge with the physical world. However, as time goes by, machine and deep learning applications are becoming more complex, requiring increasing amounts of data and training time, which makes the use of centralized approaches unsuitable. Federated learning is an emerging paradigm which enables the cooperation of edge devices to learn a shared model (while keeping private their training data), thereby abating the training time. Although federated learning is a promising technique, its implementation is difficult and brings a lot of challenges. In this paper, we present an extension of Stack4Things, a cloud platform developed in our department; leveraging its functionalities, we enabled the deployment of federated learning on edge devices without caring their heterogeneity. Experimental results show a comparison with a centralized approach and demonstrate the effectiveness of the proposed approach in terms of both training time and model accuracy.

show abstract

FedSteg: A Federated Transfer Learning Framework for Secure Image Steganalysis

Cited by 86 publications

References 37 publications

A Comprehensive Survey of Privacy-preserving Federated Learning

A Comprehensive Survey of Privacy-preserving Federated Learning

Achieving Personalized Federated Learning with Sparse Local Models

Leveraging Stack4Things for Federated Learning in Intelligent Cyber Physical Systems

Contact Info

Product

Resources

About