Abstract : Java is a cross-platform general purpose programming language. Hence, any Java based malware becomes a cross-platform threat. Since 3 Billion devices run Java, it is a serious threat. Currently, there is very little research done in the area of detection of Java malwares. As deep learning recently has proven to be effective in malware detection, we experimented with deep learning algorithms for detecting Java based malware. We name it DeepMal4J and evaluated using Long Short Term Memory (LSTM) and Gated Recurrent Unit (GRU). Our work is a first attempt to use deep neural network for the detection of Java malwares. Our system achieved accuracy of 93.33% using LSTM. This is the first ever reported results of deep learning for Java malware detection. We also present the comparison of performances and accuracy rates. Our system can be scaled up for large scale malware analysis. © Springer Nature Singapore Pte Ltd. 2019.