Federated Policy Learning for Privacy-Safe Edge Task Offloading
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional deep reinforcement learning algorithms in mobile edge computing require terminal devices to transfer private data to edge servers or remote cloud centers, risking data theft and tampering, leading to privacy and security concerns.
Innovation Solution
A policy learning method with privacy protection is developed, establishing an edge-collaborative computing offloading model and using a federated learning-based multi-agent deep reinforcement learning algorithm to autonomously learn task offloading, service caching, and transmission power control strategies, abstracting the optimization problem to a partially observable Markov decision process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional deep reinforcement learning algorithms are used for task offloading optimization, then task processing efficiency is improved, but user data privacy and security are compromised
Solution Approach 1:
The patent introduces federated learning as an intermediary mechanism between terminal devices and edge servers. Instead of directly transferring raw private data to edge servers for processing, the system uses local models at terminal devices that communicate only with each other through a centralized server. This intermediary structure enables collaborative optimization of task offloading policies while keeping private user data localized, thus resolving the contradiction between processing efficiency and data privacy protection.
Solution Approach 2:
The patent segments the centralized training process into distributed local training at multiple terminal devices. Each terminal device maintains its own local model and processes data independently, while only model parameters (not raw data) are shared with the centralized server. This segmentation approach allows the system to achieve collective intelligence for optimized task offloading without concentrating sensitive user data in a single location, thereby maintaining privacy while improving processing efficiency.
2Ease of operation
If private data is transferred to edge servers for processing, then task offloading optimization is achieved, but data security is vulnerable to third-party theft or tampering
Solution Approach 1:
The patent positions the centralized server as an intermediary that facilitates optimization without accessing raw private data. Terminal devices perform local training and only exchange encrypted model parameters with the server. This intermediary architecture enables the system to achieve optimized task offloading policies through collective learning while maintaining data security, as the server never sees or stores sensitive user information, preventing third-party theft or tampering.
Solution Approach 2:
The patent uses model parameter copies instead of raw data copies for sharing between devices. Each terminal device creates and maintains its own local model copy, and only these parameter copies are transmitted to the centralized server for aggregation. This copying approach allows the system to propagate learned optimization strategies across devices without copying or exposing actual private user data, thus achieving task offloading optimization while preserving data security and preventing third-party access to sensitive information.
Data Source
AI summary
A policy learning method with privacy protection in mobile edge computing for an intelligent agent is provided, relating to the technical field of mobile communication. The method includes: establishing an edge-collaborative computing offloading model, where the edge-collaborative computing offloading model includes a service caching model, a task offloading model, and a system cost model; establishing an optimization problem for task offloading, service caching, computing resource allocation and transmission power control based on the edge-collaborative computing offloading model for minimizing task processing costs; abstracting the optimization problem to a partially observable Markov decision process; and autonomously learning a task offloading strategy, a service caching strategy, a computing resource allocation strategy, and a transmission power control strategy by using a federated learning-based multi-agent deep reinforcement learning algorithm based on the Markov decision process.