Using Reinforcement Learning in order to detect anomalies and maybe a future response The dataset used is NSL-KDD with data of multiple anomalies

Using deep Q-Learning with keras/tensorflow to generate the network

Code associated with the paper: "Adversarial environment reinforcement learning algorithm for intrusion detection", G Caminero, M Lopez-Martin, B Carro, Computer Networks, 2019

• Detects normal or anomaly
• Train set in: AD.py
• Test set in: test.py

• Detects each attack in the dataset
• Train set in: multiAD.py
• Test set in: multi_test.py

• Detects only the attack type between normal, DoS, Probe, R2L, U2R

• Train set in: typeAD.py

• Test set in: type_test.py

• Train Dueling DDQN (tensorflow) in typeAD_tf.py

• Try to improve the inequality of attacks to produce better training
• Train set in: adversarialAD.py
• Test set in: adversarial_test.py

• Train-Test in: A3CtypeAD.py
• Summary in tensorboard: tensorboard --logdir=tmp

• NSL-KDD adaption: AE_RL_NSL-KDD.ipnb
• AWID adaption: AE_RL_awid.ipnb