LinksIQ：不完全频谱扫描下的鲁棒高效调制识别
LinksIQ: Robust and Efficient Modulation Recognition with Imperfect Spectrum Scans
摘要：
虽然对频谱共享的实际进展至关重要，但调制识别迄今为止一直在不切实际的假设下进行研究：（i）发射机的带宽必须单独和全部扫描，（ii）技术的先验知识必须可用，以及（iii）发射机必须值得信赖。实际上，这些假设不能容易地满足，因为发射机的带宽可能仅被间歇地、部分地或与其他发射机一起扫描，并且调制混淆可能由短暂的扫描或恶意活动引入。本文介绍了LinksIQ，它弥合了现实世界的频谱感知和简化假设下设计的不断增长的modrec方法之间的差距。我们的关键见解是，有序的智商样本在调制中形成了独特的模式，即使在扫描缺陷的情况下也会持续存在。我们通过Fisher核框架挖掘这些模式，并采用轻量级的线性支持向量机进行调制分类。LinksIQ对噪声、扫描干扰和数据偏差具有鲁棒性，无需利用发射机技术的先验知识。它的准确性始终优于模拟和真实轨迹中的基线。我们使用两个流行的SDR平台，RTL-SDR和USRP的测试床评估LinksIQ的性能。我们证明了高检测精度（即0.74），即使在50%的发射机重叠20美元的RTL-SDR扫描。这比RTL-SDR扫描上使用的现有同行平均提高了43%。我们还探讨了平台感知分类器训练的效果，并讨论了对现实世界的modrec系统设计的影响。我们的研究结果表明，低成本的发射机指纹在规模上的可行性。
英文摘要：
While critical for the practical progress of spectrum sharing, modulation recognition has so far been investigated under unrealistic assumptions: (i) a transmitter's bandwidth must be scanned alone and in full, (ii) prior knowledge of the technology must be available and (iii) a transmitter must be trustworthy. In reality these assumptions cannot be readily met, as a transmitter's bandwidth may only be scanned intermittently, partially, or alongside other transmitters, and modulation obfuscation may be introduced by short-lived scans or malicious activity. This paper presents LinksIQ, which bridges the gap between real-world spectrum sensing and the growing body of modrec methods designed under simplifying assumptions. Our key insight is that ordered IQ samples form distinctive patterns across modulations, which persist even with scan deficiencies. We mine these patterns through a Fisher Kernel framework and employ lightweight linear support vector machine for modulation classification. LinksIQ is robust to noise, scan partiality and data biases without utilizing prior knowledge of transmitter technology. Its accuracy consistently outperforms baselines in both simulated and real traces. We evaluate LinksIQ performance in a testbed using two popular SDR platforms, RTL-SDR and USRP. We demonstrate high detection accuracy (i.e. 0.74) even with a $20 RTL-SDR scanning at 50% transmitter overlap. This constitutes an average of 43% improvement over existing counterparts employed on RTL-SDR scans. We also explore the effects of platform-aware classifier training and discuss implications on real-world modrec system design. Our results demonstrate the feasibility of low-cost transmitter fingerprinting at scale.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
