Advancements in deep learning have enabled the development of effective solutions to different real-world problems. In this paper, we propose a deep learning model for malware detection in computing and network systems. The architecture of our proposed system consists of three modules, namely Image auto-encoder, Text auto-encoder, and GAN. Any incoming application should go through a proxy to create a log file. The log files that contain API call sequences are used to generate image and text representations that capture the dynamic behavior of the malware. Our system integrates the features from image and text auto-encoder into a single feature vector. The feature vector extracted is individually used to classify whether the log file is from a malware application. We have created a new dataset containing different newly generated malware of similar distribution. In the detection phase, the incoming log file is converted into a single feature vector by the same method as in generation. Based on the similarity index between the newly generated malware and incoming file, log files are classified as malware or benign. Thus, the newly generated dataset serves as a comparison base for the classification of log files. We have evaluated our system through a test dataset by comparing it with the newly generated malware.