Cost-Efficient and Multi-Functional Secure Aggregation in Large Scale Distributed Application.

Area of Science:

• Distributed systems
• Data mining
• Cryptography

Background:

• Secure aggregation is crucial for distributed applications and data mining.
• Existing schemes offer limited multi-functional statistical support, posing an open problem.
• Efficiently generating multiple statistics simultaneously without data recovery is a key requirement.

Purpose of the Study:

• To present SEDAR (SEcure Data Aggregation), a novel scheme for multi-functional secure aggregation.
• To address the limitations of existing schemes in supporting diverse statistical results.
• To reduce communication costs in secure aggregation through a range segmentation model.

Main Methods:

• Proposed a Range segmentation model to capture data characteristics and optimize aggregation strategies.
• Developed SEDAR by encoding raw data into value- and order-preserving vectors.
• Utilized homomorphic encryption for robust data privacy.
• Introduced enhanced versions: REDAR (Random based) and CEDAR (Compression based) to further reduce communication costs.

Main Results:

• SEDAR enables multi-functional aggregation with value and order preservation.
• REDAR and CEDAR offer significant communication cost reductions.
• REDAR trades lower security, while CEDAR trades lower accuracy for cost savings.
• Experimental evaluations on real data demonstrated excellent performance in terms of cost and accuracy.

Conclusions:

• SEDAR provides an effective solution for secure, multi-functional data aggregation.
• The Range segmentation model enhances efficiency and reduces communication overhead.
• REDAR and CEDAR offer flexible trade-offs between security, accuracy, and communication cost.
• The proposed schemes show strong practical applicability and performance.