In recent years the number of Internet of Things (IoT) devices have increased significantly. Since IoT devices are both constrained and used in industrial appli- cations across various fields such as agriculture and healthcare where security is important they have become target for malicious intrusions. To effectively protect IoT networks Intrusion Detection Systems (IDS) are used. A promising approach for designing IDS for IoT is Anomaly-Based IDS that uses machine learning to dif- ferentiate between normal and anomalous network traffic. The goal of this thesis is to design and implement an IDS based on Long Short-Term Memory (LSTM) that detects three variations of the Blackhole attack in IoT networks. Further- more this thesis evaluates the performance of the LSTM with different sequence sizes and bin sizes and compares the performance of the LSTM with a baseline model. The generalizability of the LSTM-based IDS is also evaluated across the three variations of the Blackhole attack and three types of network topologies. Lastly the speed in which the LSTM based IDS can detect intrusions is examined. The results show that the LSTM-based IDS is a good approach in terms of gen- eralizability, detection speed and ability to detect Blackhole attacks of all three variations in IoT networks.