This paper proposes TrustMesh, a novel blockchain-enabled framework that addresses trust challenges in distributed computing through a unique three-layer architecture combining permissioned blockchain technology with a novel multi-phase Practical Byzantine Fault Tolerance (PBFT) consensus protocol.
This chapter presents a comprehensive taxonomy of blockchain-enabled federated learning (BCFL) systems, analyzing coordination structures, consensus mechanisms, storage architectures, and trust models to demonstrate how specialized protocols can enable trustless collaborative machine learning across diverse domains.
SketchGuard scales Byzantine-robust decentralized federated learning by screening neighbors via compact Count Sketches and introducing commit-then-sketch to defeat a null-space attack, matching state-of-the-art robustness while cutting on-the-wire communication by up to 42%.
This paper presents Murmura, a DFL framework for wearable IoT that uses evidential deep learning for trust-aware model personalization. Epistemic uncertainty indicates peer compatibility, enabling nodes to exclude incompatible peers. Evaluation on three wearable datasets shows only 0.9% performance degradation under non-IID conditions.
This paper proposes a multi-layer IoT-Edge-Cloud healthcare architecture combining differential privacy with a hybrid Laplace-Gaussian noise mechanism, achieving 82-84% ML accuracy while reducing inference attacks by up to 70%, with blockchain for data integrity and edge computing delivering 8× latency reduction for emergencies.
Kafila serves LLMs on a small, trusted group of heterogeneous consumer devices by planning block placement and NAT-traversing ring order from each device’s measured bandwidth, capacity, and reachability. This cuts the pipeline’s slowest stage by 4.2x versus uniform division and enables serving models uniform division can’t fit at all.
This work identifies hierarchical federated learning’s aggregation topology as a privacy channel orthogonal to DP-SGD, and derives a topology-aware noise allocation that provably matches uniform DP-SGD’s per-client privacy guarantee while recovering up to 88% of the noise budget.
This work shows that in decentralized learning, adversaries can attack the collaboration topology itself, not just model updates, and introduces DMTT, a trust-screened aggregation protocol that provably bounds Byzantine influence while empirically matching no-attack accuracy across mobile sensing benchmarks where existing robust aggregators collapse.
This paper proposes a ‘Periodic Space’ for organizing the distributed computing landscape across two dimensions: a tiers continuum and abstraction levels. Analogous to the chemical periodic table, the framework characterizes system properties like responsiveness and elasticity, and anticipates future trends toward higher tiers and abstractions.