Clustering Learned CNN Features from Raw I/Q Data for Emitter Identification

Wong, Lauren J.; Headley, William C.; Andrews, Seth; Gerdes, Ryan; Michaels, Alan J.

doi:10.1109/milcom.2018.8599847

Cited by 68 publications

(40 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These prior deep learning-based SEI works either (i) perform analysis without any channel impairments (e.g., interference, noise, multipath) [44]- [46], [49], [50], (ii) use a noise only channel [41], [42], [47], [51], [52], (iii) use a multipath channel with unspecified or unknown characteristics [43], [48], [53], [54], or (iv) use a static multipath channel (i.e., the same multipath channel coefficients are used for every transmitted waveform) [35]. These works assume that the chosen deep learning approach will sufficiently learn the channel to mitigate its impact on the classification decisions.…”

Section: Related Workmentioning

confidence: 99%

Identification of OFDM-Based Radios Under Rayleigh Fading Using RF-DNA and Deep Learning

2021

View full text Add to dashboard Cite

The Internet of Things (IoT) is here and has permeated every aspect of our lives. A disturbing fact is that the majority of all IoT devices employ weak or no encryption at all. This coupled with recent advances within the areas of computational power and deep learning has increased interest in Specific Emitter Identification (SEI) as an effective means of IoT security. Deep learning is capable of in-situ extraction of discriminating features, making it well suited to discrimination of wireless transmitters without the need for feature engineering. However, the accuracy of the deep learning model is adversely affected by time-varying channel conditions. The time-varying nature is attributed to the mobility of the transmitter, receiver, objects within the operations environment, or combinations thereof. This can result in the channel conditions changing faster than the deep learning algorithm is capable of handling. This paper assesses deep learning-based SEI using waveforms that undergo Rayleigh fading, as well as channel estimation and equalization, prior to being input into a deep learning algorithm.

show abstract

Section: Related Workmentioning

confidence: 99%

Identification of OFDM-Based Radios Under Rayleigh Fading Using RF-DNA and Deep Learning

2021

View full text Add to dashboard Cite

show abstract

“…Prior security threats to cognitive signal classifiers have been researched [26], [27], yet, the state of the art signal classification systems use deep learning techniques [8]- [13] whose vulnerabilities have not been studied extensively in the context of RF. In [16] and [28], the authors consider adversarial machine learning for intelligently jamming a deep learning enabled transmitter, at transmission time and sensing time, to prevent a transmission.…”

Section: Related Workmentioning

confidence: 99%

Evaluating Adversarial Evasion Attacks in the Context of Wireless Communications

Flowers

Buehrer

Headley

2020

IEEE Trans.Inform.Forensic Secur.

Self Cite

106

View full text Add to dashboard Cite

Recent advancements in radio frequency machine learning (RFML) have demonstrated the use of raw in-phase and quadrature (IQ) samples for multiple spectrum sensing tasks. Yet, deep learning techniques have been shown, in other applications, to be vulnerable to adversarial machine learning (ML) techniques, which seek to craft small perturbations that are added to the input to cause a misclassification. The current work differentiates the threats that adversarial ML poses to RFML systems based on where the attack is executed from: direct access to classifier input, synchronously transmitted over the air (OTA), or asynchronously transmitted from a separate device. Additionally, the current work develops a methodology for evaluating adversarial success in the context of wireless communications, where the primary metric of interest is bit error rate and not human perception, as is the case in image recognition. The methodology is demonstrated using the well known Fast Gradient Sign Method to evaluate the vulnerabilities of raw IQ based Automatic Modulation Classification and concludes RFML is vulnerable to adversarial examples, even in OTA attacks. However, RFML domain specific receiver effects, which would be encountered in an OTA attack, can present significant impairments to adversarial evasion.

show abstract

“…After entering the era of deep learning, Xu et al [14] bring forward the three-way incremental learning algorithm for emitter identification, which is adaptive to the increase of emitter types and samples. In [15], the method applying Convolutional Neural Networks (CNN) as feature learners and extractors is proposed. However, the methods in [14] and [15] both depend on the differences of the signal modulation types or modulation parameters.…”

Section: Introductionmentioning

confidence: 99%

Dynamic Curve Fitting and BP Neural Network With Feature Extraction for Mobile Specific Emitter Identification

2021

View full text Add to dashboard Cite

As a critical technology in both civil and military fields, specific emitter identification (SEI) can identify signal sources according to their various features. Existing methods on SEI are mostly based on the prior knowledge of emitters, which are powerless in the non-cooperative scenario. In order to realize the unsupervised identification, the mobile SEI method based on fingerprint set construction and feedback classification algorithm is proposed in this paper. The proposed method first divides signal fingerprints into static features and dynamic features, where the former describe the inherent features of emitters, and the latter represent the moving state features of emitters. Then, the feedback classification algorithm composed of dynamic curve fitting and back propagation (BP) neural network is applied in the classification of signals.The dynamic curve accomplishes the first classification and the results with high credibility are used to train the BP neural network which accomplishes the final classification. Simulation results demonstrate that the proposed method can complete the identification of mobile specific emitter sources in the unsupervised state with more than 95% identification rate.INDEX TERMS SEI, curve fitting, signal fingerprint, BP neural network.

show abstract

Clustering Learned CNN Features from Raw I/Q Data for Emitter Identification

Cited by 68 publications

References 15 publications

Identification of OFDM-Based Radios Under Rayleigh Fading Using RF-DNA and Deep Learning

Identification of OFDM-Based Radios Under Rayleigh Fading Using RF-DNA and Deep Learning

Evaluating Adversarial Evasion Attacks in the Context of Wireless Communications

Dynamic Curve Fitting and BP Neural Network With Feature Extraction for Mobile Specific Emitter Identification

Contact Info

Product

Resources

About