Baseline result is much better than state-of-the-art model

I am researching about Deep Learning based Intrusion Detection System. I found a paper on a well-known journal, which is considered as a state-of-the-art method in this research area, because it got many citations. In the paper, they proposed using Inception Resnet v4 to solve the problem and got the lowest error rate, compared to other studies.

I am developing a new method using their data pre-processing idea. First, I built a baseline, which is a very simple and shallow CNN. However, it produces a very good result, which is much better than the study using Inception Resnet (approximately 1-2%).

Note that: we are using the same dataset as well as data preprocessing as they described in the paper.

In addition, I am worry about that there maybe some mistakes while preprocessing data. Therefore, I re-implemented their proposed network and run it. The re-implemented results is not much different to the reported results in paper. So I think, my preprocessing step is correct.

The paper did not compare their results with a simple CNN model. They just compared with other architectures like LSTM, ANN, and basic machine learning models.

I am quite confused now. What should I do now?