Reimplementation of AiDroid (AiDroid: When Heterogeneous Information Network Marries Deep Neural Network for Real-time Android Malware Detection).
Disclaimer: This is not official code repository, and I am not responsible for the accuracy of the code.
dataset description: There are 6 types of nodes (i.e., apk, api(dense), permission, tpl, url, component) in HIN and 4 metapaths (i.e., apk-tpl-apk, apk-permission-apk, apk-api-apk, apk-component-apk, apk-url-apk) are consider in this repository.
run scripts: (your data_path needs to specificed in each script before runing)
# step1: biased random walk. get corpus from HIN
python randomwalk.py
# step2: learn representation for each node
python node2vec.py
# step3: hin2Img.
python hin2img.py
# step4: train classifer.
python train.py