Secure URLLC Design for Task Offloading in UAV Edge Computing Systems

Introduction

The rapid development of the Internet of Things (IoT) has led to the emergence of numerous mobile applications in smart homes, smart cities, healthcare, and precision agriculture. However, the increasing number of connected devices generates massive amounts of data, posing significant challenges for IoT devices with limited computational capabilities and energy resources. Mobile edge computing (MEC) has emerged as a promising solution, allowing users to offload computationally intensive tasks to edge servers, thereby alleviating the computational burden on local devices.

Despite its advantages, MEC faces challenges in remote or harsh environments where ground users may lack access to edge computing servers. Unmanned aerial vehicles (UAVs) offer a flexible and cost-effective solution by carrying MEC servers to extend coverage in such areas. UAV-assisted MEC systems can provide on-demand computational support, particularly in scenarios requiring high-speed, low-latency data transmission.

Ultra-reliable and low-latency communication (URLLC) is critical for real-time applications, demanding extremely high reliability (decoding error probability ≤ 10⁻⁵) and minimal latency (≤ 1 ms for small data packets). However, URLLC systems operate with short block lengths to reduce latency, leading to non-zero decoding errors and information leakage risks. Traditional encryption-based security methods are unsuitable for URLLC due to the short block lengths, making physical layer security a more viable alternative.

This paper addresses the physical layer security challenges in UAV-assisted MEC systems where multiple ground users offload tasks to a UAV equipped with an MEC server while a ground eavesdropper attempts to intercept the transmitted data. The primary objective is to maximize the minimum secure computational capacity among users by jointly optimizing the UAV’s deployment position, task offloading bandwidth allocation, and CPU frequency allocation for both local and remote computations.

System Model

Communication Model

The system consists of multiple ground users randomly distributed in a designated area, a UAV equipped with an MEC server hovering at a fixed altitude, and a ground-based eavesdropper. The UAV assists users in task computation while the eavesdropper attempts to intercept the transmitted data.

Given the need for low-latency communication, users transmit data in short packets, ensuring transmission within the channel coherence time. Frequency-division multiple access (FDMA) is employed to allow simultaneous transmissions without interference. The channel between users and the UAV follows a line-of-sight (LoS) dominant free-space path loss model, while the eavesdropper’s channel is modeled using Rayleigh fading.

The achievable secure transmission rate under URLLC constraints is derived, accounting for decoding errors and information leakage probabilities. The rate expression incorporates channel dispersion effects, which characterize the randomness of the channel compared to deterministic channels.

Computation Model

The computation process involves three stages: local computation at the user, task offloading to the UAV, and remote computation at the UAV. Each user’s computational task is bit-independent, and the required CPU cycles per bit are defined.

Local Computation: Users perform partial computations locally, consuming energy proportional to the CPU frequency and computation time. The local computation must adhere to energy constraints to prevent excessive power consumption.
Remote Computation: The UAV receives offloaded tasks and processes them using allocated CPU resources. The remote computation must satisfy causality constraints, ensuring that only successfully received data is computed.

The total secure computational capacity for each user is the sum of locally computed and remotely computed data, subject to constraints on total latency, energy consumption, bandwidth allocation, and CPU frequency limits.

Problem Formulation

The optimization problem aims to maximize the minimum secure computational capacity among users by jointly optimizing:
• UAV deployment position

• Task offloading bandwidth allocation

• Local and remote CPU frequencies

• Computation time allocation

The problem is constrained by:
• Maximum latency requirements

• Energy consumption limits

• Bandwidth and CPU frequency constraints

• Information causality for remote computation

Due to the non-convex nature of the problem, a block coordinate descent (BCD) approach is adopted, decomposing the problem into two subproblems:

Subproblem 1: Optimizing UAV position and computation time given fixed bandwidth and CPU frequencies.
Subproblem 2: Optimizing bandwidth and CPU frequencies given fixed UAV position and computation time.

These subproblems are solved iteratively until convergence.

Solution Methodology

Optimization of UAV Position and Computation Time

Given fixed bandwidth and CPU frequencies, the UAV position and computation time are optimized to maximize the minimum secure computational capacity. The non-convex constraints are addressed using logarithmic function approximations and successive convex approximation (SCA) techniques.

The secure transmission rate is lower-bounded using a log-function approximation, and the resulting convex problem is solved using standard optimization tools.

Optimization of Bandwidth and CPU Frequencies

With fixed UAV position and computation time, the bandwidth and CPU frequencies are optimized. The non-convex rate expressions are linearized using first-order Taylor expansions, transforming the problem into a convex form.

The optimal bandwidth allocation is derived using Lambert W functions, and CPU frequencies are optimized under energy constraints. A two-layer algorithm is employed, where the inner layer optimizes bandwidth and CPU frequencies for a fixed objective, and the outer layer maximizes the minimum secure computational capacity.

Alternating Optimization Algorithm

The two subproblems are solved alternately until convergence. The algorithm’s complexity is analyzed, demonstrating efficient convergence with manageable computational overhead.

Simulation Results

Simulations are conducted to evaluate the proposed algorithm under different eavesdropper positions, bandwidth allocations, and CPU frequency constraints. Key findings include:

UAV Positioning: The UAV adjusts its position based on the eavesdropper’s location to enhance secure communication. When the eavesdropper is near a specific user, the UAV moves closer to that user to mitigate interception risks.
Bandwidth Allocation: Users closer to the eavesdropper receive more bandwidth to compensate for higher security risks. As the eavesdropper moves toward the center, bandwidth allocation becomes more balanced.
CPU Frequency Allocation: Higher UAV CPU frequencies improve computational capacity until bandwidth limitations become the bottleneck.
Performance Comparison: The proposed joint optimization scheme outperforms benchmark approaches, including:
• UAV position optimization only

• Bandwidth and CPU optimization only

• Shannon capacity-based approximations

The results validate the necessity of joint optimization in balancing secure communication and computational efficiency.

Conclusion

This paper presents a comprehensive framework for secure URLLC in UAV-assisted MEC systems, addressing physical layer security challenges while ensuring low-latency and reliable task offloading. By jointly optimizing UAV deployment, bandwidth allocation, and computational resource management, the proposed algorithm maximizes the minimum secure computational capacity among users.

Simulation results demonstrate significant performance improvements over existing schemes, highlighting the effectiveness of the proposed approach. The study provides valuable insights into achieving secure and efficient UAV-assisted MEC systems, paving the way for future research in URLLC-enabled edge computing.

doi.org/10.19734/j.issn.1001-3695.2024.05.0180

Was this helpful?

0 / 0