Detecting malware is an essential component of cybersecurity to safeguard systems from malicious threats. An innovative hybrid model termed Long Short Feed Forward Base Convolution Neural Network (LSFFbCNN) is proposed for accurate and efficient malware detection. It combines Long Short-Term Memory (LSTM) networks and Feedforward Neural Networks (FFNN) to boost malware detection capabilities. The proposed method seeks to enhance the accuracy and efficiency of malware detection by exploiting the advantage of both sequential and non-sequential feature extraction techniques. The distinct feature vectors extracted from these networks are integrated through a concatenation layer, creating a unified data representation. This combined feature vector is further refined by a dense layer that increases the model’s ability to absorb information from sequential and non-sequential features. It employs CNNs as classifiers, utilizing Reinforcement Learning (RL) and SoftMax functions to identify malware patterns. The proposed LSFFbCNN is evaluated on a ‘Malware detection’ dataset that includes malware and benign samples, demonstrating superior accuracy and robustness in distinguishing malicious software. The proposed LSFFbCNN achieved 99.98% accuracy and 99.97% precision, including F1-score and recall. The results highlight the efficacy of this hybrid approach, presenting a promising solution for strengthening malware detection in cybersecurity applications.