Safeguarding Cluster Heads in UAV Swarm Using Edge Intelligence: Linear Discriminant Analysis-Based Cross-Layer Authentication

Huanchi Wang,Xianbin Wang,He Fang

doi:10.1109/ojcoms.2021.3084532

Huanchi Wang, Xianbin Wang + Show 1 more

Open Access

https://doi.org/10.1109/ojcoms.2021.3084532

Copy DOI

Abstract

While the unmanned aerial vehicles (UAVs) swarm travels under a dynamic environment, the cluster head (CH) switching is unavoidable due to the mitigation of mobility, quality of service, and energy consumption. If an attacker becomes the new CH, the entire swarm will be controlled and the sensitive data will be leaked. Unlike the other mobile networks with constant network connectivity, the authentication in the UAV swarm suffers from intermittent connection with the ground station under a hostile environment or spectrum constraint condition. Hence, this paper proposes a novel CH safeguarding mechanism enabled by edge intelligence utilizing a situational-aware authentication scheme. This low-latency mechanism provides extra security at the CH selection and switching without cloud server support. By adopting the unique cross-layer attributes, the system security is significantly improved based on the extracted multi-dimensional information. The Linear Discriminant Analysis (LDA) algorithm fuses the authentication decision accurately by projecting the high dimensional estimations into a low dimensional space for maximum separability by only keeping the necessary attributes. A situation-aware cross-layer attribute selection algorithm is developed to select a minimum number of attributes so that the time required for attribute estimation and computation overhead of authentication can be reduced. The simulation results demonstrate that our scheme performs better under a dynamic environment compared with the physical layer authentication scheme and some existing state-of-the-art authentication techniques.

Highlights

T HE UNMANNED Aerial Vehicles (UAVs) have become ubiquitous in both civilian and military fields in recent years due to the flexibility in operation and risk reduction of personal injury [1]
An edge intelligence-enabled safeguarding mechanism has been proposed in this paper to enhance the security in the unmanned aerial vehicles (UAVs) swarm
Network by forging and deleting the data packets. This multi-dimensional authentication scheme was planted in the on-duty cluster head (CH) to verify that the candidate UAVs are legitimate before the CH selection process and the new CH is still legitimate before the CH switching process

Summary

INTRODUCTION

T HE UNMANNED Aerial Vehicles (UAVs) have become ubiquitous in both civilian and military fields in recent years due to the flexibility in operation and risk reduction of personal injury [1]. The rapid computational power development increases the potential of deciphering the encryption algorithms, which makes the attacker easier to impersonate a legitimate UAV and become the new CH [18] Another popular approach in UAV authentication enhancement is to utilize the physical layer attributes, i.e., the radio frequency (RF) fingerprinting, which has been used widely in intrusion detection, access control, cloning detection and malfunction detection [19], [20]. A potential approach is to use the sequential forward selection mechanism in which attributes are added sequentially until the criterion of selection has been reached [37] This technique can select the minimum amount of attributes that are most relevant to the environment; the process of hillclimbing search requires a high computational power that is not suitable in our edge network [38].

SYSTEM MODEL AND PROBLEM FORMULATION

AUTHENTICATION BASED ON LDA ALGORITHM

PERFORMANCE EVALUATION

Findings

CONCLUSION