Comparisons of attacks on honeypots with those on real networks

Abstract

Honeypots are computer systems deliberately designed to be attack targets, mainly to learn about cyber-attacks and attacker behavior. When implemented as part of a security posture, honeypots also protect real networks by acting as a decoy, deliberately confusing potential attackers as to the real data. The objective of this research is to compare attack patterns against a honeypot to those against a real network, the network of the Naval Postgraduate School. Collection of suspicious-event data required the implementation and setup of a honeypot, in addition to the installation and use of an intrusion-detection system. A statistical analysis was conducted across suspicious-event data recorded from a honeypot and from a real network. Metrics used in our study were applied to the alerts generated from Snort 2.4.3, an open-source intrusion detection system. Results showed differences between the honeypot and the real network data which need further experiments to understand. Both the honeypot and the real network data showed much variability at the start of the experiment period and then a decrease in the number of alerts in the later period of the experiment. We conclude that after the initial probing and reconnaissance is complete, the vulnerabilities of the network are learned and therefore fewer alerts occur; but more specific signatures are then aimed at exploiting the network.