quebic-source / ml-vulnerability-scanner

Convolution neural network based ml model to predict vulnerability type of the given source code

Prerequisite

• java
• ant
• python
• docker
• joern-0.3.1 installation

Install project dependencies

• cd project
• sudo pip install -r requirements.txt

Run neo4j on docker

• sudo docker run -p 7474:7474 -p 7687:7687 --mount type=bind,source=,target=/conf neo4j:2.3.0 dump-config
• sudo docker run -p 7474:7474 -p 7687:7687 --mount type=bind,source=,target=/data --mount type=bind,source=,target=/conf neo4j:2.3.0
• open neo4j on browser localhost:7474

Config neo4j

• Jump into local config dir
• neo4j.properties -> allow_store_upgrade=true
• neo4j-server.properties -> org.neo4j.server.webserver.address=0.0.0.0

Run

• training sudo python app/main.py --training
• testing sudo python app/main.py --testing

Tensorboard

• tensorboard --logdir /home/tharanga/tensorflow_logs