Information-Theoretically Secure Aggregation for Lightweight Federated Learning: Resilient to Dropouts and Adversaries

2026-07-23 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy · Depth: Expert, quick

Summary

A new framework proposes information-theoretically secure aggregation for lightweight federated learning (FL), specifically targeting sign-based methods on resource-constrained devices. While sign-based approaches reduce communication costs, they are vulnerable to inference attacks, and existing secure aggregation schemes often incur significant overhead or incompatibility. This framework securely computes the majority vote (MV) polynomial via single-round secure multiplication, ensuring end-to-end information-theoretic security under an honest-majority assumption, revealing only the final aggregated sign. It employs inverse-form exponent reduction, halving the MV polynomial degree to reduce costs, and single-round secure multiplication for linear offline complexity and single online communication. These techniques reduce online communication by up to 99.5% and latency by up to 85.7%. Leveraging MDS-code-based decoding, the framework is robust against dropouts and adversarial behaviors, yielding accuracy gains of up to 20.65% and 10.74% respectively, establishing a practical foundation for large-scale, low-latency secure aggregation in sign-based FL.

Key takeaway

For AI Scientists and Machine Learning Engineers designing secure and efficient federated learning systems for resource-constrained devices, this framework offers a practical path to deploy sign-based FL with strong privacy guarantees and high performance. You should consider integrating inverse-form exponent reduction and MDS-code-based decoding to achieve significant reductions in communication (up to 99.5%) and latency (up to 85.7%), while improving model accuracy and resilience against dropouts and adversaries.

Key insights

Secure aggregation for sign-based FL can be lightweight and information-theoretically robust against dropouts and adversaries.

Principles

• Information-theoretic security is achievable for sign-based FL.
• Reducing polynomial degree significantly cuts communication and computation.
• MDS-code-based decoding enhances FL robustness.

Method

Securely compute the majority vote polynomial using single-round secure multiplication, enhanced by inverse-form exponent reduction and MDS-code-based decoding for efficiency and robustness.

In practice

• Implement single-round secure multiplication for FL aggregation.
• Apply inverse-form exponent reduction to optimize MV polynomial degree.
• Utilize MDS codes for dropout and adversarial resilience in FL.

Topics

• Federated Learning
• Secure Aggregation
• Information-Theoretic Security
• Sign-based Gradients
• MDS Codes
• Communication Efficiency

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, AI Security Engineer

Related on AIssential

• Privacy-Preserving and Verifiable Approximate Distributed Coded Computing — Model-agnostic framework unifies privacy and adversarial resilience in distributed learning via coded computing and specific defenses.
• FedIDM: Achieving Fast and Stable Convergence in Byzantine Federated Learning through Iterative Distribution Matching — FedIDM uses distribution matching and condensed data to achieve fast, stable, and robust federated learning convergence.
• FedCVESA: Taking Away Training Data in Federated Learning via Correlation Value Encoding and Segmented Aggregation — Federated learning can be exploited by malicious servers to embed and extract private training data.
• Decentralized Federated Learning by Partial Message Exchange — PaME improves DFL communication, privacy, and accuracy by exchanging sparse, randomly selected model coordinates.
• Entropy-Guided Tensor Compression for Multimodal Federated Learning on Edge Devices — Spectral entropy guides adaptive compression rank allocation in multimodal federated learning, improving efficiency and accuracy on heterogeneous edge devices.
• No More Guessing: a Verifiable Gradient Inversion Attack in Federated Learning — VGIA offers verifiable gradient inversion for tabular data, ensuring exact record reconstruction with certified correctness.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.