Cyberattacks have become a significant danger to digital data and infrastructure as digital network usage rises rapidly and reliance on technology spreads over all spheres. From distributed denial of service (DDoS) attacks to brute force attacks to long-term targeted attacks (APTs), these attacks call for creating intelligent and strong systems to identify these hazards properly. A hybrid model using LSTM and CNN models with an attention mechanism to improve the performance of detecting cyberattacks has been introduced. CNN model aims to extract spatial features from network data, and LSTM is used to check temporal patterns associated with attack activities. Attention mechanism has a significant role in the proposed model in classifying difficult attacks by improving the model's accuracy, which enhances concentration on the key aspects of the input data. The proposed hybrid model (CNN-LSTM-Attention mechanism) attained an accuracy of up to 98% despite significant difficulties in defining overlapping patterns. The results revealed the model's ability to reduce errors and increase attack detection. The proposed study indicates that increasing performance by employing more varied data and creating analytical methods involves attention to multi-head. This model is developed towards more intelligent and effective cyber security solutions.