In the development of Internet of Things (IoT), network security has received increasing attention. Many network attacks are performed through the sensors. Among the various cyberattacks, distributed denial-of-service (DDoS) attacks represent one of the most serious types. DDoS attacks should not be underestimated in terms of the loss they may cause. In this study, we integrated Apache Spark, a big-data computing framework, with a detection model based on a back-propagation artificial neural network. Thanks to the capacity of a big-data computing framework for mass historical data, computation of the characteristics required for the learning of the detection model can be performed in a real-time manner. An artificial neural network model is perfect for DDoS detection, owing to its good scalability and its advantage of restricting the expansion of computing resources consumed as data volume increases. This eliminates the problems related to traditional approaches to DDoS signature computing, where mass data can take considerable time to compute and can even overwhelm a system. The results of this study show that the trained artificial neural network achieved a detection rate as high as 99.80% and the real-time detection system achieved a detection rate of 87.18%. Compared with other studies in this field, it is clear that the proposed approach provides effective detection of DDoS attacks, and that the incorporation of Apache Spark, the open-source clustering big-data computing framework, allows more computers and more Kafka Producers receiving packets to be used to form an even larger detection system capable of dealing with increasingly huge and diverse DDoS attacks.