Creating Sequential GAN

[ommprakash....@itbhu.ac.in]

Hello All,

I am Omm prakash sahoo, a 3rd student from IIT BHU and a Data Science BS student from IIT Madras.

Currently, we are working on a project, i.e., on Detecting Insider Treats from publicly available data CERT insider threat dataset.

Since the CERT insider threat dataset r4.2 contains more insider threat instances than other versions of the dataset, we choose it as the experimental dataset. The dataset captures 32,770,227 events (log lines) of 1,000 users (with 70 insiders) over 17 months, which consists of five types of event records, including logon, device, email, file, and HTTP. Among these are 7,323 malicious events generated by 70 insiders, representing three insider threat scenarios (Information leak, Data theft, and IT sabotage).

Due to this highly unbalanced data distribution, we want to create a GAN model to augment the malicious data to a reasonable size. So, please help us if there is any seqGAN code available to solve the issue. 

Thank you,

Omm prakash