Corrupted cert count dataset

[AlkaSaliss]

Hello, I was trying ot reproduce the baseline presented in the paper "Deep Learning for Unsupervised Insider Threat Detection in Structured Cybersecurity Data Streams"
However the dataset containing count features provided in this repository seems to be corrupted, some rows are mixed, and some columns are shifted
could you point to the script used to produce the data, or to a non-corrupted dataset ?
Thanks