by Vit Bukac, April 2014, 83 pages.
FIMU-RS-2014-02. Available as Postscript, PDF.
Denial of service (DoS) attacks is an ever growing threat to the availability of computer systems. Numerous solutions have been proposed both for DoS attacks detection and mitigation. However, their evaluation and mutual comparison is complicated due to scarcity of representative contemporary input data. In academia, proposed DoS detection systems are frequently evaluated with obsolete and in practice no longer used tools. Such discrepancy can lead to distinctly different detection efficiency in evaluation environment and real environment. To address this issue, we provide a comparative analysis of traffic features of DoS attacks that were generated by state-of-the-art standalone DoS attack tools. We list frequently used traffic features and verify their presence in analyzed attack traffic. Common denominator of all attack traffic is the presence of repeated similar yet independent operations. Therefore, we propose a new research area for the detection of DoS attacks the source end, based on repeated attack patterns recognition.