一种木马图片检测方法、系统及计算机可读存储介质

• Format: Patent
• Language: Chinese
• Published: 25.02.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

本发明涉及一种木马图片检测方法、系统及计算机可读存储介质；木马图片检测方法包括：S1、对图像数据集中的图片进行解码、切分；其中，图像数据集包括正常图片和木马图片；S2、根据切分得到的文本样本构建文本总体特征；S3、对文本向量分别通过加入注意力的双向LSTM和三个不同kernel_size的并行CNN进行特征提取，得到LSTM特征和CNN特征；S4、将三种特征组合，之后依次通过全连接层和softmax激活函数，建立混合神经网络模型并进行权重训练；S5、将待检测图片进行解码、切分，然后经过处理得到三种特征并输入权重训练之后的混合神经网络模型，输出待检测图片的分类结果。本发明提升了木马检测的精度。 The invention relates to a Trojan horse picture detection method and system and a computer readable storage medium. The Trojan horse picture detection method comprises the following steps: S1, decoding and segmenting pictures in an image data set; wherein the image data set comprises normal pictures and Trojan pictures; S2, constructing text overall features according to the text samples obtained through segmentation; S3, performing feature extraction on the text vector through a bidirectional LSTM with attention added and a parallel CNN with three different kernel sizes, and obtaining LSTM features and CNN features; S4, the three features are combined, then a hybrid neural network model is established through a fu