Scarce Attack Datasets and Experimental Dataset Generation

Efficient employability of machine learning and artificial intelligence techniques to automate the process of network forensics for attack identification and attribution, depends largely on the availability of datasets to build up the knowledge base. The requirement in the case of a more proactive approach of incident response and mitigation is also the same. This paper explore the publicly available attack datasets with regard to their employability in ML and AI for network forensics and consider the need for developing ethical hacking techniques for generation of attack datasets through experimentation. We assess the methodology with certain attack scenarios and observe that the same can be adopted for a wide gamut of network attacks. The findings are potentially valuable for researchers and security practitioners to overcome the limitations of the scarce attack datasets.