Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Adversarial Threats to Automatic Modulation Open Set Recognition in Wireless Networks

Yandie Yang, Sicheng Zhang, Kuixian Li, Qiao Tian, Yun Lin

TL;DR

This work investigates adversarial threats to Automatic Modulation Open Set Recognition (AMOSR) in wireless networks by introducing OSAttack, a framework comprising OSFGSM and OSPGD to attack both discriminative (OpenMax, CenterLoss, CGDL) and generative (OpenGAN) AMOSR models. Through experiments on the RADIOML2016.10A dataset, the study shows that small adversarial perturbations substantially degrade unknown-sample rejection, as evidenced by sharp drops in AUS and AUROC, with CenterLoss and CGDL particularly vulnerable. The results underscore the vulnerability of AMOSR to carefully crafted adversarial examples and highlight the need for robust defenses and transferability-aware attack strategies in spectrum monitoring and interference management systems.

Abstract

Automatic Modulation Open Set Recognition (AMOSR) is a crucial technological approach for cognitive radio communications, wireless spectrum management, and interference monitoring within wireless networks. Numerous studies have shown that AMR is highly susceptible to minimal perturbations carefully designed by malicious attackers, leading to misclassification of signals. However, the adversarial security issue of AMOSR has not yet been explored. This paper adopts the perspective of attackers and proposes an Open Set Adversarial Attack (OSAttack), aiming at investigating the adversarial vulnerabilities of various AMOSR methods. Initially, an adversarial threat model for AMOSR scenarios is established. Subsequently, by analyzing the decision criteria of both discriminative and generative open set recognition, OSFGSM and OSPGD are proposed to reduce the performance of AMOSR. Finally, the influence of OSAttack on AMOSR is evaluated utilizing a range of qualitative and quantitative indicators. The results indicate that despite the increased resistance of AMOSR models to conventional interference signals, they remain vulnerable to attacks by adversarial examples.

Adversarial Threats to Automatic Modulation Open Set Recognition in Wireless Networks

TL;DR

This work investigates adversarial threats to Automatic Modulation Open Set Recognition (AMOSR) in wireless networks by introducing OSAttack, a framework comprising OSFGSM and OSPGD to attack both discriminative (OpenMax, CenterLoss, CGDL) and generative (OpenGAN) AMOSR models. Through experiments on the RADIOML2016.10A dataset, the study shows that small adversarial perturbations substantially degrade unknown-sample rejection, as evidenced by sharp drops in AUS and AUROC, with CenterLoss and CGDL particularly vulnerable. The results underscore the vulnerability of AMOSR to carefully crafted adversarial examples and highlight the need for robust defenses and transferability-aware attack strategies in spectrum monitoring and interference management systems.

Abstract

Automatic Modulation Open Set Recognition (AMOSR) is a crucial technological approach for cognitive radio communications, wireless spectrum management, and interference monitoring within wireless networks. Numerous studies have shown that AMR is highly susceptible to minimal perturbations carefully designed by malicious attackers, leading to misclassification of signals. However, the adversarial security issue of AMOSR has not yet been explored. This paper adopts the perspective of attackers and proposes an Open Set Adversarial Attack (OSAttack), aiming at investigating the adversarial vulnerabilities of various AMOSR methods. Initially, an adversarial threat model for AMOSR scenarios is established. Subsequently, by analyzing the decision criteria of both discriminative and generative open set recognition, OSFGSM and OSPGD are proposed to reduce the performance of AMOSR. Finally, the influence of OSAttack on AMOSR is evaluated utilizing a range of qualitative and quantitative indicators. The results indicate that despite the increased resistance of AMOSR models to conventional interference signals, they remain vulnerable to attacks by adversarial examples.
Paper Structure (17 sections, 22 equations, 6 figures, 1 table)

This paper contains 17 sections, 22 equations, 6 figures, 1 table.

Table of Contents

  1. Introduction
  2. Background and Threat model
  3. Background of Adversarial Attack
  4. Adversarial Threat Model of AMOSR
  5. Attack Methods and Measurement Indicator
  6. OSAttack of Discriminative Model
  7. OpenMax and CenterLoss
  8. CGDL
  9. OSAttack of Generative Model
  10. Evaluation of Indicators
  11. Experiments and Discussions
  12. Dataset
  13. Analysis of AUS
  14. Analysis of AUROC
  15. Analysis of Probability Distribution
  16. ...and 2 more sections

Figures (6)

  • Figure 1: The OSAttack framework diagram for OpenMax and CenterLoss.
  • Figure 2: The OSAttack framework diagram for CGDL.
  • Figure 3: The OSAttack framework diagram for OpenGAN.
  • Figure 4: The AUS of different AMOSR models under OSFGSM and OSPGD attacks as the perturbation changes.
  • Figure 5: The AUROC of different AMOSR models under OSFGSM and OSPGD attacks as the perturbation changes.
  • ...and 1 more figures