This repository show the notes about papers I read and the papers are almost about malware detection
这个仓库里是我读paper时的笔记,其中的论文大多是关于恶意软件检测。
2017-07-05 Deep Android Malware Detection - Niall McLaughlin, etc
- paper title: Deep Android Malware Detection
- paper authors: Niall McLaughlin, Jesus Mar tinez del Rincon, etc
- paper keywords: Malware Detection, Android, Deep Learning
这篇论文使用静态操作码作为特征,CNN为模型,在实验环境下,F-sorce=0.97,在实际环境中,F-sorce=0.71。 虽然实际环境中的效果不是特别好,但是模型训练所需要的时间相对较少,而且内存使用率上是一个常数。
2017-05-24 Malware Detection with Deep Neural Network Usinig Process Behavior - Shun T., etc
- paper title: Malware Detection with Deep Neural Network Usinig Process Behavior
- paper authors: Shun Tobiyama, Yukiko Yamaguchi, Hajime Shimada, etc
- paper keywords: malware infection detection, neural network, process behavior
这篇论文以进程行为作为源数据,之后使用RNN从源数据中抽取特征,并将抽取到的特征转化成特征图像, 之后使用CNN以特征图像作为输入进行分类。使用ROC曲线作为评价标准,在最好的情况下AUC=0.96。
2017-04-19 Malware Classification with Recurrent Networks - Rasvan P., etc
- paper title: Malware Classification with Recurrent Networks
- paper authors: Rasvan Pascanu, Jack W.Stokes, Hermineh Sanossian, Mady Marinescu, Anil Thomas
- paper keywords: Malware Classification, Recurrent Neural Network, Deep Learning
这篇论文使用ESN作为循环模型进行特征抽取,Max-Pooling做非线性抽样,logistic回归做最后的分类;这样的混合模型达到最好的效果。 相对标准的混合三元模型,本文提出的这个混合模型的true positive rate=98.3% & false positive rate=0.1%。
2017-04-12 Efficient Dynamic Malware Analysis Based on Network Behavior Using Deep Learning - Toshiki S., etc
- paper title: Efficient Dynamic Malware Analysis Based on Network Behavior Using Deep Learning
- paper authors: Toshiki Shibahara, Taakeshi Yagi, Mitsuaki Akiyama, etc
这篇论文基于网络行为来确定是否应暂停动态分析的方法,使用RNN模型,在对29,562个恶意软件样本的评估中,提出的方法将分析时间缩短了67.1%, Precision=0.976, Recall=0.962, F-measure=0.969。
2017-04-07 A multi-task learning model for malware classification with useful file access pattern from API call sequence - Xin W., etc
- paper title: A multi-task learning model for malware classification with useful file access pattern from API call sequence
- paper authors: Xin Wang, Siu Ming Yiu
这篇论文