Deep Learning for RF Signal Classification in Unknown and Dynamic Spectrum Environments

Shi, Yi; Davaslıoğlu, Kemal; Sagduyu, Yalin E.; Headley, William C.; Fowler, Michael C.; Green, Gilbert

doi:10.1109/dyspan.2019.8935684

Cited by 94 publications

(45 citation statements)

References 24 publications

(25 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One example of DL model in wireless domain is modulation recognition to classify signals into modulation types. Beyond traditional approaches that use carefully designed features (cyclic spectrum) [10], [11], recent efforts have applied the I/Q samples directly as input to a CNN [1], [12], [13].…”

Section: Related Workmentioning

confidence: 99%

Trojan Attacks on Wireless Signal Classification with Adversarial Machine Learning

Davaslıoğlu

Sagduyu

2019

2019 IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN)

Self Cite

View full text Add to dashboard Cite

We present a Trojan (backdoor or trapdoor) attack that targets deep learning applications in wireless communications. A deep learning classifier is considered to classify wireless signals using raw (I/Q) samples as features and modulation types as labels. An adversary slightly manipulates training data by inserting Trojans (i.e., triggers) to only few training data samples by modifying their phases and changing the labels of these samples to a target label. This poisoned training data is used to train the deep learning classifier. In test (inference) time, an adversary transmits signals with the same phase shift that was added as a trigger during training. While the receiver can accurately classify clean (unpoisoned) signals without triggers, it cannot reliably classify signals poisoned with triggers. This stealth attack remains hidden until activated by poisoned inputs (Trojans) to bypass a signal classifier (e.g., for authentication). We show that this attack is successful over different channel conditions and cannot be mitigated by simply preprocessing the training and test data with random phase variations. To detect this attack, activation based outlier detection is considered with statistical as well as clustering techniques. We show that the latter one can detect Trojan attacks even if few samples are poisoned.

show abstract

Section: Related Workmentioning

confidence: 99%

Trojan Attacks on Wireless Signal Classification with Adversarial Machine Learning

Davaslıoğlu

Sagduyu

2019

2019 IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN)

Self Cite

View full text Add to dashboard Cite

show abstract

“…RF signal classification can support different applications such as radio fingerprinting [28] that can be ultimately used in cognitive radio systems [29] subject to dynamic and unknown interference and jamming effects [30]. Modulation classification has been extensively studied with deep neural networks [6], [7], [20]- [27], where the goal is to classify a given signal to a known modulation type. Different types of datasets have been used to train deep neural network for modulation classification.…”

Section: Related Workmentioning

confidence: 99%

Real-Time and Embedded Deep Learning on FPGA for RF Signal Classification

Soltani

Sagduyu

Hasan

et al. 2019

MILCOM 2019 - 2019 IEEE Military Communications Conference (MILCOM)

Self Cite

View full text Add to dashboard Cite

We designed and implemented a deep learning based RF signal classifier on the Field Programmable Gate Array (FPGA) of an embedded software-defined radio platform, DeepRadio TM , that classifies the signals received through the RF front end to different modulation types in real time and with low power. This classifier implementation successfully captures complex characteristics of wireless signals to serve critical applications in wireless security and communications systems such as identifying spoofing signals in signal authentication systems, detecting target emitters and jammers in electronic warfare (EW) applications, discriminating primary and secondary users in cognitive radio networks, interference hunting, and adaptive modulation. Empowered by low-power and low-latency embedded computing, the deep neural network runs directly on the FPGA fabric of DeepRadio TM , while maintaining classifier accuracy close to the software performance. We evaluated the performance when another SDR (USRP) transmits signals with different modulation types at different power levels and DeepRadio TM receives the signals and classifies them in real time on its FPGA. A smartphone with a mobile app is connected to DeepRadio TM to initiate the experiment and visualize the classification results. With real radio transmissions over the air, we show that the classifier implemented on DeepRadio TM achieves high accuracy with low latency (microsecond per sample) and low energy consumption (microJoule per sample), and this performance is not matched by other embedded platforms such as embedded graphics processing unit (GPU).

show abstract

“…Recently, deep learning techniques have been rapidly developed and have made great strides in the signal identification field. For example, convolutional neural networks can be said to be the most popular architecture for both modulation and wireless technology recognition [ 15 , 16 ]. Although this approach performs well in different applications and has the advantage of simple feature pre-processing or even raw data usage, it requires large-scale training data resulting in high implementation costs [ 17 ].…”

Section: Introductionmentioning

confidence: 99%

V2X Wireless Technology Identification Using Time–Frequency Analysis and Random Forest Classifier

Skiribou

Elbahhar

2021

Sensors

View full text Add to dashboard Cite

Signal identification is of great interest for various applications such as spectrum sharing and interference management. A typical signal identification system can be divided into two steps. A feature vector is first extracted from the received signal, then a decision is made by a classification algorithm according to its observed values. Some existing techniques show good performance but they are either sensitive to noise level or have high computational complexity. In this paper, a machine learning algorithm is proposed for the identification of vehicular communication signals. The feature vector is made up of Instantaneous Frequency (IF) resulting from time–frequency (TF) analysis. Its dimension is then reduced using the Singular Value Decomposition (SVD) technique, before being fed into a Random Forest classifier. Simulation results show the relevance and the low complexity of IF features compared to existing cyclostationarity-based ones. Furthermore, we found that the same accuracy can be maintained regardless of the noise level. The proposed framework thus provides a more accurate, robust and less complex V2X signal identification system.

show abstract

Deep Learning for RF Signal Classification in Unknown and Dynamic Spectrum Environments

Abstract: Dynamic spectrum access (

Cited by 94 publications

References 24 publications

Trojan Attacks on Wireless Signal Classification with Adversarial Machine Learning

Trojan Attacks on Wireless Signal Classification with Adversarial Machine Learning

Real-Time and Embedded Deep Learning on FPGA for RF Signal Classification

V2X Wireless Technology Identification Using Time–Frequency Analysis and Random Forest Classifier

Contact Info

Product

Resources

About