Small versions of the extracted features datasets for 9 attacks on IP camera and IoT networks generated by Mirskey et al (2018).

Published: 17-11-2018| Version 1 | DOI: 10.17632/zvsk3k9cf2.1
Ran Yahalom,
Alon Steren,
Yonatan Nameri,
Max Roytman


Mirsky et al. (2018) recorded traffic from two networks: (1) an operational IP camera video surveillance network on which they conducted 8 attacks that affect the availability and integrity of the video uplinks; (2) a noisier IoT network consisting of 3 PCs and 9 IoT devices, one of which was infected with the Mirai botnet malware. A detailed description of the attacks and network typologies can be found in their article (see attached link): Y. Mirsky, T. Doitshman, Y. Elovici, and A. Shabtai, “Kitsune: An Ensemble of Autoencoders for Online Network Intrusion Detection,” in NDSS, 2018. For each of these 9 attacks, Mirsky et al. compiled a dataset of extracted feature vectors that were extracted for every packet. These datasets and their descriptions are available from the authors at attached link. From each of these datasets, we extracted a small segment of consecutive packets which we split into training and test sets as described in the included "Dataset Properties.csv". Using these smaller datasets allows the quick evaluation of attack detection experiments. This can be useful when you need to conduct an exhaustive yet relatively fast model selection for a network intrusion detection system (IDS) given a large parameter value space. It also convenient for testing & debugging during IDS development.