.png){kind=icon}
Return
Resource-Aware Secure Multi-Party Edge Computation Offloading
.png){kind=icon}
PRE
AI
DOI：10.1109/tcomm.2026.3714302.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Computation offloading enables resource-limited users to delegate resource-intensive tasks to more powerful edge devices (workers), to reduce the computational load on the user. A key challenge is ensuring the privacy of sensitive data when interacting with untrusted devices, while simultaneously adapting the offloading mechanism under resource heterogeneities. To address this challenge, this work proposes a resource-aware secure computation offloading framework, where we provide information-theoretic privacy guarantees for both the sensitive user data and the computation results, while adapting the secure offloading mechanism to the communication and computation resource availabilities of the workers. We leverage an offline-online communication paradigm where communication-intensive randomness generation operations are carried out in a task-independent offline phase, and propose a neural multi-armed bandit mechanism to minimize online system latency under resource heterogeneities. Our theoretical analysis presents formal privacy and regret guarantees, while extensive evaluations demonstrate that the proposed framework achieves up to $5.9\times $ faster online task completion time compared to conventional baselines and $12\times $ faster processing time compared to performing the computations locally at the user.
Keywords:
Secure computation offloading
coded computing
information-theoretic privacy
neural bandits
Journal
IF:
8.3
Papers:
1.2W
Citations:
3.6W
Organization
Cited Papers
.png){kind=icon}
ShareSaveJoint Parameter-and-Bandwidth Allocation for Improving the Efficiency of Partitioned Edge Learning
OAAI
OAAIShareSave
ShareSaveShareSaveShareSaveShareSaveShareSaveShareSave
View more