Internet of Things (IoT) is expanding rapidly as its use cases and technology advances. However, along with these benefits, IoT also faces many attacks. One of the most common and difficult-to-detect attacks is the DDoS attack and its variations (MrDDoS). Implementing security solutions in the IoT ecosystem is challenging due to resource-constrained devices. The proposed security solution, namely, MAD-HOT utilized the Hoeffding tree and placement strategies for MrDDoS attack detection with reasonable accuracy. The placement problem is formulated with a hypergraph and a game theory algorithm. Security module based on Hoeffding trees is the main classifier, along with many other classifiers for model training. Finally, for prediction, we have used logistic regression as the meta-classifier. Extensive experimentation on the Contiki cooja simulator shows that the MAD-HOT can best execute the balance between MrDDoS detection and energy overhead.