Enhancing Split Learning with Sharded and Blockchain-Enabled SplitFed Approaches

TL;DR

The paper tackles scalability and security in SplitFed Learning by introducing two frameworks: Sharded SplitFed Learning (SSFL), which distributes the SL server workload across multiple shards with a federated aggregation layer to stabilize updates, and Blockchain-enabled SplitFed Learning (BSFL), which replaces the centralized server with a committee-based blockchain consensus to evaluate and aggregate shard updates. SSFL significantly improves scalability and convergence speed, while BSFL adds resilience to data poisoning and reduces the risks associated with central server failures. The authors validate both approaches on Fashion MNIST, showing that BSFL maintains performance and dramatically improves robustness under attack, with SSFL offering substantial gains in throughput and stability. This work advances privacy-preserving distributed learning by delivering end-to-end decentralized, secure, and scalable SplitFed architectures suitable for sensitive-data domains.

Abstract

Collaborative and distributed learning techniques, such as Federated Learning (FL) and Split Learning (SL), hold significant promise for leveraging sensitive data in privacy-critical domains. However, FL and SL suffer from key limitations -- FL imposes substantial computational demands on clients, while SL leads to prolonged training times. To overcome these challenges, SplitFed Learning (SFL) was introduced as a hybrid approach that combines the strengths of FL and SL. Despite its advantages, SFL inherits scalability, performance, and security issues from SL. In this paper, we propose two novel frameworks: Sharded SplitFed Learning (SSFL) and Blockchain-enabled SplitFed Learning (BSFL). SSFL addresses the scalability and performance constraints of SFL by distributing the workload and communication overhead of the SL server across multiple parallel shards. Building upon SSFL, BSFL replaces the centralized server with a blockchain-based architecture that employs a committee-driven consensus mechanism to enhance fairness and security. BSFL incorporates an evaluation mechanism to exclude poisoned or tampered model updates, thereby mitigating data poisoning and model integrity attacks. Experimental evaluations against baseline SL and SFL approaches show that SSFL improves performance and scalability by 31.2% and 85.2%, respectively. Furthermore, BSFL increases resilience to data poisoning attacks by 62.7% while maintaining superior performance under normal operating conditions. To the best of our knowledge, BSFL is the first blockchain-enabled framework to implement an end-to-end decentralized SplitFed Learning system.

Figures (4)

• Figure 1: Overview of Sharding SplitFed Learning framework.
• Figure 2: Performance comparison for 9 nodes in 60 training rounds.
• Figure 3: Performance comparison for 36 nodes in 30 training rounds.
• Figure 4: Round completion times for 36 nodes.