ScaleSFL

Madill, Evan; Nguyen, Ben; Leung, Carson K.; Rouhani, Sara

doi:10.1145/3494106.3528680

Cited by 20 publications

(2 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Madill et al. [20] proposed ScaleSFL, an extensible block‐chain‐based partitioning solution for federal learning, which uses sharding as well as partitioning‐level and main‐chain consensus for federal training. Each shard is much smaller than the overall size.…”

Section: Background and Related Workmentioning

confidence: 99%

Enhancing the blockchain interoperability through federated learning with directed acyclic graph

Xia,

Kaiye,

Songze

et al. 2023

IET Blockchain

View full text Add to dashboard Cite

The use of federated learning to achieve blockchain interoperability has become a hot topic in research, because it enables data exchange without revealing any private information. However, the previous work, such as ScaleSFL (Asia‐CCS, 2022), that has implemented federated learning for blockchain interoperability, the throughput of the framework still cannot support the practical applications. Therefore, a federated learning framework based on Directed Acyclic Graph (DAG) is proposed which utilizes sharding mechanism to enhance the blockchain interoperability. By constructing a weighted context graph based on data correlation, reasonable sharding of the dataset is achieved, thereby improving the efficiency of blockchain interoperability. The experimental results show that the federated framework reduces global computation in federated learning by 30% compared with other schemes, while increasing blockchain throughput by nearly 40%.

show abstract

Section: Background and Related Workmentioning

confidence: 99%

Enhancing the blockchain interoperability through federated learning with directed acyclic graph

Xia,

Kaiye,

Songze

et al. 2023

IET Blockchain

View full text Add to dashboard Cite

show abstract

“…For advances, intelligence is injected to sharding scheme. Yang sharding mechanism together with an extended committee consensus method and applys federated learning to dynamical sharding scheme [76]. In their design, offchain federated learning is executed for model update verification, which is beneficial for pluggable poisoning mitigation, the shard-level training provides models to be aggregated globally, and a proof-of-concept prototype is designed for consensus purpose.…”

Section: Security Solution In Bc-based Vanetsmentioning

confidence: 99%

Secured Scalable Blockchain Networks for Trustworthy Distributed Deep Learning in VANETs

View full text Add to dashboard Cite

Distributed deep learning (DDL) within vehicular ad hoc networks (VANETs) holds profound significance in developing smart applications, such as intelligent transport systems and autonomous driving, where multiple parties are coordinated to leverage the training capability and acquisitive intelligence. Blockchain (BC) is a distribution technology promising for trustworthy DDL, holding the divide-and-conquer concept with decentralized management and consensus algorithms to reduce the exposure of sensitive controllers and thus the risks of malicious system-wide attacks. Moreover, zero trust architecture (ZTA) concepts are promoted as innovative cybersecurity solution which can be integrated with BC to address the resource-limited and infrastructure-less issues to augment the strength of DDL in VANETs. In this dissertation, the BC and ZTA potential is explored to construct reliable VANETs for trustworthy data sharing and thus to support significant within-VANET DDL and relevant applications. Firstly, virtualized distributed ledger technology (vDLT) is developed as the multimedia BC platform, followed by vDLT-based VANETs built to solve the unsteady communication; secondly, vDLT is improved to transmit and secure traffic events in VANETs; subsequently, a vDLT-based DDL system is proposed to enhance object detection (OD) inside VANETs; finally, ZTA and sharding scheme are enabled in vDLT-based VANETs for improved protection. Generally, vDLT runs with virtualized resource and sharding supports for scalability improvement, along with multi-layered consensus algorithm and an adaptable hierarchical and decentralized PBAC (hdPBAC) access control i model to agilely protect the DDL procedures and relevant DDL-related applications.The proposed system has been developed, including the vDLT system, the multimedia streaming over vDLT in fixed networks and VANETs, a precursory vDLT-based DDL system for OD purpose, and the integration of ZTA into scalable-BC-based VANETs.The evaluation results show the feasibility of proposed design and demonstrate the significance of performance improvements. First and foremost, I would like to express my deepest appreaciation to my supervisor, Prof. Richard Yu, who have been working with his wholehearted dedication to supervise my PhD study. His immense knowledge and cutting-edge insights have been encouraging me of elevating my academic research quality. He have also endearvored to find fundings to support my study and daily life to keep my work proceeding smoothly. Without his invaluable advice and continous supports, the achievements during my study would have been impossible. Words are powerless to express my gratitude to him, and I will maintain my grateful heart constantly and sincerely. I am also grateful to Prof. Omair Shafiq for his valuable comments and suggestions on my research and his knowledge and experience in engineering shared for my improvements. Additionally, I would like to thank Prof. Ashraf Matrawy, Prof. Rong Liu, and Prof. Azzedine Boukerche for serving in my dissertation ...

show abstract