This paper presents a novel IoT-based intrusion detection system (IDS) using a hybrid deep learning algorithm. The study aims to detect cyber attacks, particularly DDoS attacks, in IoT networks, which are vulnerable to such threats due to their limited energy and computational resources. The IDS is designed to analyze network traffic data in a big data environment, leveraging the capabilities of Apache Spark and Google Colab. The algorithm combines one-dimensional Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks to enhance detection accuracy. The CICIoT2023 and TON_IoT datasets are used for training and testing, with feature reduction techniques applied to ensure significant feature inclusion. The proposed model is evaluated using accuracy, precision, recall, and F1-score parameters, achieving high accuracy rates of 99.995% for binary classification and 99.96% for multiclassification in the CICIoT2023 dataset, and 98.75% for binary classification in the TON_IoT dataset. The study compares the developed model with ten traditional machine learning and deep learning algorithms, demonstrating its superior performance in detecting DDoS attacks.This paper presents a novel IoT-based intrusion detection system (IDS) using a hybrid deep learning algorithm. The study aims to detect cyber attacks, particularly DDoS attacks, in IoT networks, which are vulnerable to such threats due to their limited energy and computational resources. The IDS is designed to analyze network traffic data in a big data environment, leveraging the capabilities of Apache Spark and Google Colab. The algorithm combines one-dimensional Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks to enhance detection accuracy. The CICIoT2023 and TON_IoT datasets are used for training and testing, with feature reduction techniques applied to ensure significant feature inclusion. The proposed model is evaluated using accuracy, precision, recall, and F1-score parameters, achieving high accuracy rates of 99.995% for binary classification and 99.96% for multiclassification in the CICIoT2023 dataset, and 98.75% for binary classification in the TON_IoT dataset. The study compares the developed model with ten traditional machine learning and deep learning algorithms, demonstrating its superior performance in detecting DDoS attacks.