In the Internet of Things (IoT), devices have limited computational capabilities in order to reduce their size and energy consumption, making them vulnerable to the deployment of botnets and their use in Distributed Denial of Service (DDoS) cyberattacks on remote servers, which limits the services available to legitimate users and causes economic losses. Although classical Deep Learning techniques can detect these cyberattacks, they require large amounts of data and computational resources. Quantum Machine Learning (QML) emerges as a potential solution for solving problems more energy-efficiently. This work explores the use of quantum models for classification, and in particular Quantum Support Vector Machines (QSVM) with different schemes to encode classical data (angle embedding, amplitude embedding and ZZ feature map) and detect DDoS attacks from TCP/IP packet flows. The results show that QSVMs achieve an f1-score close to classical models, with angle embedding being the best one, although it requires longer computation time due to slower convergence.