用深度学习提升物联网安全，精准识别异常行为
Enhancing Cybersecurity in IoT Networks: A Deep Learning Approach to Anomaly Detection
- 结合LSTM与注意力机制，捕捉网络流量的时序特征
- 在多个真实数据集上准确率超基线模型，最高达98.7%
- 适合网络安全工程师和物联网系统开发者参考
随着互联网与智能设备的普及，物联网技术在智能家居、城市安防、智慧物流等领域得到广泛应用。物联网支持关键生产指标的实时监控，帮助企业发现潜在质量缺陷、预测设备故障并优化流程，降低损失与成本；同时实现资产实时追踪，提升利用效率。然而，物联网设备数量激增也带来网络安全威胁上升，成为恶意攻击的重要入口。本文提出一种融合LSTM与注意力机制的深度学习模型，用于检测物联网网络中的异常行为。实验基于IoT-23、BoT-IoT、IoT network intrusion、MQTT、MQTTset等数据集，结果表明该方法在多个评估指标上优于现有基准模型。
原文摘要 · Abstract (English)
With the proliferation of the Internet and smart devices, IoT technology has seen significant advancements and has become an integral component of smart homes, urban security, smart logistics, and other sectors. IoT facilitates real-time monitoring of critical production indicators, enabling businesses to detect potential quality issues, anticipate equipment malfunctions, and refine processes, thereby minimizing losses and reducing costs. Furthermore, IoT enhances real-time asset tracking, optimizing asset utilization and management. However, the expansion of IoT has also led to a rise in cybercrimes, with devices increasingly serving as vectors for malicious attacks. As the number of IoT devices grows, there is an urgent need for robust network security measures to counter these escalating threats. This paper introduces a deep learning model incorporating LSTM and attention mechanisms, a pivotal strategy in combating cybercrime in IoT networks. Our experiments, conducted on datasets including IoT-23, BoT-IoT, IoT network intrusion, MQTT, and MQTTset, demonstrate that our proposed method outperforms existing baselines.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。