With the advancement of technology, Internet of Things (IoT) devices are integrated with smart homes, smart cities, intelligent medical systems, industries, smart cars, and many more applications to monitor and control them. These devices are connected to different heterogeneous environments and have many environmental constraints such as power, bandwidth, resources, etc. Unfortunately, this makes them attractive targets for attackers to perform malicious behaviours. It necessitates updating the current intrusion detection system (IDS) to cope with existing and new challenges. This paper proposes an IDS to secure a Message Queuing Telemetry Transport (MQTT)-based IoT environment. The MQTT does not use robust encryption algorithms to encrypt the transmitted data for fast communication, which makes the networks vulnerable to intruders and network attacks. The proposed model selects essential features using a genetic algorithm, and these selected features are used to train a convolutional neural network model for network packet classification. We have used the MQTT-IoT-IDS2020 dataset to analyze and measure the model’s performance. The test results are promising and prove that the proposed scheme can identify potential intrusions in the MQTT networks.

随着技术的进步，物联网（IoT）设备与智能家居、智能城市、智能医疗系统、工业、智能汽车以及更多应用集成以对其进行监视和控制。这些设备连接到不同的异构环境，并具有许多环境限制，例如功率、带宽、资源等。不幸的是，这使得它们成为攻击者执行恶意行为的有吸引力的目标。需要更新当前的入侵检测系统（IDS）以应对现有的和新的挑战。本文提出了一种 IDS 来保护基于消息队列遥测传输 (MQTT) 的物联网环境。 MQTT 没有使用强大的加密算法来加密传输的数据以实现快速通信，这使得网络容易受到入侵者和网络攻击。所提出的模型使用遗传算法选择基本特征，并且这些选择的特征用于训练用于网络数据包分类的卷积神经网络模型。我们使用 MQTT-IoT-IDS2020 数据集来分析和测量模型的性能。测试结果令人鼓舞，证明所提出的方案可以识别 MQTT 网络中的潜在入侵。