FPGA/AI-Powered Architecture for Anomaly Network Intrusion Detection Systems

Round 1

Reviewer 1 Report

The paper is well written and clearly presents the idea of using FPGA based anomaly detection neural networks architecture to prevent network attacks. FPGAs have proven to be better for the machine learning and deep leaning applications as compare to ASICs or GPUs, therefore it seems right approach. 

 

The results are clearly explained with details about the throughput and bandwidth analysis which are required for FPGA based architecture. It also compares the results with other existing models which is really valuable.

 

One of the concerns is that the total number of packets for training in datasets is less for a deep learning application. Is it possible that model built is an overfit and giving high results?

 

Can authors combine all the data set and test the approach and add results in the paper?

 

Can we improve the accuracy of the model by involving humans in the loop (as a feedback) to tell them which one is an intrusion and which one is not?

 

Overall paper looks good.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 2 Report

Dear Authors

The authors propose a deep learning mode for network anomaly detection. and compare by building two models to validate the proposed framework for anomaly detection that is Anomaly Detection Autoencoder (ADA) and Artificial Neural 3 Classification (ANC) in the NetFPGA-sume platform. The paper presents an interesting outcome. The paper is good and can be of substantial value for improving network intrusion detection systems. 

Author Response

First of all, we would like to thank the reviewer for your valuable feedback. The positive feedback persuades us to improve our research results in the future. We hope soon we can show more interesting results to the NIDS research society.

Many thanks and everything goes well with you.