| 引用本文： |
-
叶莺,徐锦龙,李瑞,魏冬.基于碎片极化指纹融合分析的跳频辐射源识别[J].信息安全学报,已采用 [点击复制]
- YE Ying,XU Jinlong,LI Rui,WEI Dong.Frequency Hopping Emitter Identification Based on Polarization Fingerprint Fragment Fusion Analysis *[J].Journal of Cyber Security,Accept [点击复制]
|
|
| 摘要: |
| 辐射源识别在军用领域有广泛的应用，实现辐射源识别较为成熟的技术是射频指纹。然而，随着制造工艺的提升，射频指纹存在指纹差异小、低信噪比环境下鲁棒性较差的缺点,使其在同型号设备的识别中应用受限。近期提出的极化指纹由于同时具有频率相关性和方向相关性，可以有效扩大设备指纹间的差异。但现有基于极化指纹的辐射源识别研究中多局限于极化指纹的低维特征，对其高维特征未能充分挖掘。此外，由于跳频通信频率的时变性，跳频信号指纹在频域存在碎片化问题，其在目前研究中尚未得到讨论。为解决上述不足，本文提出一种基于极化指纹和深度学习的跳频辐射源识别方案，通过挖掘跳频设备极化指纹的高维数据特征，并将投票机制引入深度学习来减少指纹碎片化带来的影响，提高方案的识别性能和抗噪声性能。实验分别分析了信噪比、指纹碎片数和指纹碎片分布对PF-CNN(Polarization fingerprint convolutional neural network)性能的影响。结果表明，在同等信噪比条件下，本文所提方案与其他特定辐射源识别方案相比性能最佳。当信噪比小于10dB时，PF-CNN识别准确率可以达到98.7%。 |
| 关键词: 极化指纹 射频指纹 特定辐射源识别 卷积神经网络 |
| DOI： |
| 投稿时间：2023-02-19修订日期：2023-04-18 |
| 基金项目: |
|
| Frequency Hopping Emitter Identification Based on Polarization Fingerprint Fragment Fusion Analysis * |
|
YE Ying1,2, XU Jinlong1,2, LI Rui1,2, WEI Dong1,2
|
| (1.School of Cyber Security,University of Chinese Academy of Sciences,Beijing;2.Institute of Information Engineering, Chinese Academy of Sciences) |
| Abstract: |
| Specific emitter identification has a wide range of applications in the military field, and the more mature technology to achieve specific emitter identification is radio frequency fingerprint. However, with the manufacturing process im-provement, RF fingerprint has the disadvantages of small fingerprint difference and poor robustness in low SNR en-vironments, so its application in the identification of the same type of equipment is limited. The recently proposed polarization fingerprint can effectively expand the difference between device fingerprint due to both frequency and directional dependence. However, the existing research on specific emitter identification based on polarization finger-print is mostly limited to the low-dimensional features of polarization fingerprint, and the high-dimensional features are not fully explored. In addition, due to the time-varying nature of frequency hopping communication frequency, there is a fragmentation problem of frequency hopping signal fingerprint in the frequency domain, which has not been discussed in the current research. To solve the above shortcomings, this paper proposes a frequency-hopping emitter identification scheme based on polarization fingerprint and deep learning. By mining the high-dimensional data fea-tures of polarization fingerprint of frequency hopping devices and introducing the voting mechanism into deep learning to reduce the impact of fingerprint fragmentation, the identification performance and anti-noise performance of the scheme are improved. The effects of signal-to-noise ratio, the number of fingerprint fragments, and fingerprint frag-mentation distribution on the performance of PF-CNN (Polarization fingerprint convolutional neural network) are an-alyzed separately. The results show that the proposed scheme has the best performance compared with other specific emitter identification schemes under the same signal-to-noise ratio. When the signal-to-noise ratio is less than 10 dB, the identification accuracy of PF-CNN can reach 98.7%. |
| Key words: polarization fingerprint radio frequency fingerprint specific emitter identification convolutional neural network |
