Enhancing Split Learning with Sharded and Blockchain-Enabled SplitFed Approaches

2025-09-29

AI Summary

Existing federated learning (FL) incurs high computational overhead, while split learning (SL) suffers from significant training latency; although SplitFed (SFL) integrates both paradigms, it remains hindered by poor scalability, suboptimal performance, and weak security guarantees. This paper proposes a sharded, blockchain-empowered SplitFed framework. We design a novel sharded parallel architecture to offload server-side computation and establish an end-to-end decentralized system. A committee-based consensus mechanism and model-update evaluation strategy are introduced to ensure training fairness and robustness against adversarial attacks. Furthermore, a multi-layer defense mechanism is integrated to mitigate data poisoning and model integrity attacks. Experimental results demonstrate that our framework improves training performance by 31.2%, scalability by 85.2%, and resilience to data poisoning by 62.7% over baseline methods, while maintaining superior convergence behavior and model accuracy under benign conditions.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Natural Language Processing: Safety and Robustness
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Security and Privacy: Blockchains and distributed ledgers
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

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.

Problem

• Addressing scalability and performance limitations in SplitFed Learning
• Enhancing security against data poisoning and model integrity attacks
• Reducing computational demands and training times in distributed learning

Innovation

• Sharded SplitFed distributes workload across parallel shards
• Blockchain-enabled SplitFed uses committee consensus for security
• Evaluation mechanism excludes poisoned model updates effectively