Low-Latency and Secure Computation Offloading Assisted by Hybrid\n Relay-Reflecting Intelligent Surface

Khac–Hoang Ngo, Nhan T. Nguyen, Thinh Quang Dinh, Trong-Minh Hoang, Markku Juntti

TL;DR

This paper introduces a Hybrid Relay-Reflecting Intelligent Surface (HRRIS) to enhance Mobile Edge Computing (MEC). By integrating a few active amplifying elements into a mostly passive reflecting surface, the authors demonstrate a significant reduction in computation offloading latency while maintaining data secrecy against eavesdroppers, consistently outperforming traditional passive RIS.

What Problem Does This Paper Address?

Mobile Edge Computing (MEC) allows resource-constrained devices to offload heavy computational tasks to nearby edge servers. However, this process faces two major hurdles:

  1. Latency: Wireless channels are prone to fading and attenuation. If the connection is poor, offloading can actually take longer than processing the task locally.
  2. Security: Offloading data over the air makes it vulnerable to malicious eavesdroppers.

While Reconfigurable Intelligent Surfaces (RIS) have been proposed to "steer" radio waves and improve these links, they are purely passive. They cannot amplify signals, which limits their effectiveness when the signal is already weak. The authors address these gaps by proposing the HRRIS—a surface that can both reflect and amplify signals—to minimize latency under strict security constraints.

Methods

The authors model a system consisting of a single user, an Edge Node (EN), a multi-antenna eavesdropper, and an HRRIS. Their approach involves a joint design of communication and computation parameters:

  • The HRRIS Architecture: Unlike a standard RIS, the HRRIS contains $N$ elements where a small subset ($A$) are active (equipped with power amplifiers) and the rest are passive.
  • Alternating Optimization (AO): Because the problem of minimizing latency while maximizing the "secrecy rate" is non-convex and complex, the authors use an AO algorithm. It iteratively solves for:
  • Offloading Volume: Determining how many bits to process locally vs. at the edge.
  • Receive Combining: Optimizing how the Edge Node combines incoming signals to maximize the Signal-to-Interference-plus-Noise Ratio (SINR).
  • HRRIS Coefficients: Optimizing the phase shifts (for all elements) and amplification factors (for active elements).
  • Dynamic vs. Fixed HRRIS: The paper evaluates two modes. In "Fixed HRRIS," the active elements are in set positions. In "Dynamic HRRIS," the system can choose which elements to activate based on real-time channel conditions.

Datasets

The paper relies on numerical simulations rather than physical datasets. The authors use standard wireless channel models:

  • Large-scale fading: Modeled using standard path-loss exponents.
  • Small-scale fading: Modeled using Rician fading (accounting for both Line-of-Sight and non-Line-of-Sight components).
  • Parameters: Simulations typically assume an Edge Node with 5 antennas, an HRRIS with 50 elements (often with only 1 active element), and specific coordinate-based locations for the user and eavesdropper.

Key Insights and Ideas

  • The Power of One: One of the most striking findings is that adding even a single active element to a passive surface of 50 elements provides a massive boost in performance. The ability to amplify the signal overcomes the "double path-loss" effect that plagues traditional RIS.
  • Dynamic Flexibility: Dynamic HRRIS (where the active elements can change position) significantly outperforms fixed architectures. This is because it can adapt to the specific spatial geometry of the user and the eavesdropper.
  • Security-Latency Synergy: The HRRIS doesn't just make the signal stronger; it makes the legitimate signal stronger relative to what the eavesdropper hears. This increases the "secrecy rate," allowing more data to be offloaded securely rather than being forced into slower local processing.

Key Results

  • Latency Reduction: HRRIS consistently achieves lower latency than both local computing and conventional passive RIS.
  • Scaling with Surface Size: As the number of surface elements ($N$) increases, the gain of fixed HRRIS over optimized passive RIS eventually shrinks. However, Dynamic HRRIS maintains a high gain even as the surface grows.
  • Robustness to Eavesdropping: In scenarios where an eavesdropper has many antennas (making them very "capable"), traditional RIS and local computing fail to provide low latency. In contrast, a Dynamic HRRIS with 10dBm of power can still reduce latency by approximately 9% compared to other schemes.

Limitations and Future Work

  • Single-User Focus: The current model only considers one user. In a real-world cell, multiple users would compete for the HRRIS's reflection beams and the Edge Node's computing cycles.
  • CSI Assumptions: The study assumes the Edge Node has perfect Channel State Information (CSI) for the legitimate links, which is difficult to achieve in practice with high-frequency (mmWave) channels.
  • Hardware Complexity: While HRRIS is more efficient than a full-blown relay, the active elements require a power supply and more complex hardware than passive RIS, which may impact deployment costs.
  • Future Directions: The authors suggest extending this work to multi-user systems and exploring the impact of resource allocation between competing users.

Summary updated

Abstract

Recently, the hybrid relay-reflecting intelligent surface (HRRIS) has been\nintroduced as a spectral- and energy-efficient architecture to assist wireless\ncommunication systems. In the HRRIS, a single or few active relay elements are\ndeployed along with a large number of passive reflecting elements, allowing it\nto not only reflect but also amplify the incident signals. In this work, we\ninvestigate the potential of the HRRIS in aiding the computation offloading in\na single-user mobile edge computing system. The objective is to minimize the\noffloading latency while ensuring the secrecy of user data against a malicious\neavesdropper. We develop efficient solutions to this latency minimization\nproblem based on alternating optimization. Through numerical results, we show\nthat the deployment of the HRRIS can result in a considerable reduction in\nlatency. Furthermore, the latency reduction gain offered by the HRRIS is much\nmore significant than that of the conventional reconfigurable intelligent\nsurface (RIS).\n
BibTeX
@misc{ngo2021low,
  author = {Khac–Hoang Ngo and Nhan T. Nguyen and Thinh Quang Dinh and Trong-Minh Hoang and Markku Juntti},
  title = {{Low-Latency and Secure Computation Offloading Assisted by Hybrid\textbackslash\{\}n Relay-Reflecting Intelligent Surface}},
  year = {2021},
  eprint = {2109.01335},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2109.01335},
}
Figure