A review on lexical based malicious domain name detection methods

Nowadays, domain names are becoming crucial digital assets for any business. However, the media never stopped reporting phishing and identity theft attacks held by third-party entities that rely on domain names to mislead Internet users. Thus, Palo Alto Networks revealed in their studies 20 largely cyber-squatted domain names targeting popular brands. Based on their behavior, domain names appear in public lists that objectively evaluate their reputation. Blacklists contain domain names that have previously committed suspicious acts, whereas whitelists include the most popular and trustworthy domain names. For a long time, this listing technique has been used as a reactive approach to counter domain name-based attacks. However, it suffers from the limitation of responding late to attacks. Nowadays, techniques tend to be much more proactive. They operate before any attack occurs. As part of the CSNET conference, we published a short paper that describes a plethora of domain name attacks and their associated detection techniques using their lexical features (Hamroun et al. 2022). In this paper, we present an extended version of the original one which discusses the previously mentioned points in more detail and adds some elements of understanding when it comes to malicious domain name detection. Hence, we provide a literature review of malicious domain name detection techniques that use only the lexical features of domain names. These features are available, privacy-preserving, and highly improve detection results. The review covers recent works that report relevant performance categorized according to a new taxonomy. Moreover, we introduce a new criterion for comparing all the existing works based on targeted maliciousness type before discussing the limitations and the newly emerging research directions in this field.