结合注意力与双线性网络的Android恶意软件检测

Combining attention and bilinear networks for Android malware detection

为解决基于字节码图像的Android恶意软件检测方法中存在多分类准确率不高、易被噪音数据干扰等问题,提出一种结合注意力与双线性网络的恶意软件检测方法。从软件样本中提取字节码文件并将其映射为彩色图像,将注意力机制加入双线性网络的一条分支中减小噪音数据的影响,使用结合注意力的双线性网络模型完成对字节码图像的分类。实验结果表明,该方法在恶意软件二分类与多分类中均具有良好的性能,总体检测性能优于其它同类方法与模型。

To solve the problems of multi-classification accuracy and interference by noisy data in bytecode images based Android malware detection method,a malware detection method combining attention and bilinear network was proposed.Bytecode files were extracted from software samples and they were mapped to color images,the attention mechanism was added to one branch of the bilinear network to reduce the influence of noisy data,and the bilinear network model combined with attention was used to complete the classification of bytecode images.Experimental results show that the method has good performance in both malware binary classification and multi-classification,and the overall detection performance is better than other similar methods.