The volume and complexity of log data have increased exponentially due to the rapid proliferation of modern software systems, making anomaly detection a crucial task for maintaining system security and dependability. By combining attention processes with Bidirectional Long Short-Term Memory (Bi-LSTM) networks, this research introduces a novel method for software log anomaly identification. The proposed model leverages the sequential dependencies in log data captured by Bi-LSTM, while the attention mechanism enhances the interpretability and focuses on key features indicative of anomalies. We compare the model’s performance against state-of-the-art techniques in terms of detection accuracy, precision, recall, and F1-score using publically accessible log datasets. The results highlight the robustness of the proposed approach in identifying subtle and complex anomalies, offering a scalable solution for real-world applications. This study not only advances anomaly detection capabilities in software systems but also provides insights into the synergistic effects of deep learning architectures and attention mechanisms for log analysis.