Intrusion Detection Systems (IDS) are important for securing networks against ever-evolving cyber threats. This paper proposes a leveraging model based on Long Short-Term Memory (LSTM) capability to model sequential dependencies in network traffic data for an anomaly-based IDS capable of capturing temporal patterns, the system efficiently identifies anomalies that evolve, outperforming traditional machine learning models. This proposed model TimeSeq-IDS involves processing network traffic logs, training an LSTM model for classifying data as normal or anomalous and deploying real-time detection mechanisms. Using comparative analysis with existing models, including Support Vector Machines (SVM), Random Forest (RF), and Convolutional Neural Networks (CNN), demonstrates the superior performance of the LSTM model to identify sophisticated attack patterns with lower false positive rates and high accuracy. The proposed model, TimeSeq-IDS, achieves a precision of 98%, recall value of 97% and F1 Score of 97.5%. Moreover, the paper estimates the system’s efficiency in handling large-scale datasets, such as CICIDS 2017, and provides insights into its real-world applications, including critical infrastructure and enterprise networks. The proposed framework shows enhanced scalability, adaptability and detection accuracy, making it a promising solution for modern cyber security challenges such as sequential modelling capability, temporal pattern recognition, real-time threat detection with low latency, higher detection accuracy and robustness against imbalanced data. By combining temporal sequence learning, adaptive anomaly detection, and real-time threat intelligence, TimeSeq-IDS offers superior accuracy, scalability and resilience against emerging cyber threats – making it an optimal choice for modern enterprise, IoT, and cloud security environments.