The importance of automated vulnerability analysis techniques is growing as more software is developed. In this research, we present a deep learning-based method for learning assembly code in order to detect software flaws. Unlike previous research that relied on API function call sequences, our method begins by storing the assembly code in an immutable vector before using deep learning to learn the assembly language. When it comes to modeling assembly code, we choose Instruction2vec, which is efficient in vectorizing the code. We classify if the new functions have software weaknesses or not after learning the assembly code of the current functions using the vector provided by Instruction2vec. Many ways to detecting vulnerabilities using deep learning have been developed to solve vulnerabilities. Most learning-based approaches, on the other hand, discover vulnerabilities in source code rather than binary code. We present our method for detecting vulnerabilities in binary code in this paper. Our method builds deep learning models to discover vulnerabilities using binary code produced from the SARD dataset.