Against Credential Thief - A Modular Honeytoken Based Framework

Sophisticated hacking techniques pose significant risks to both corporate and personal data, creating a formidable challenge for traditional defense methods. Cyber deception is considered a promising mechanism that can effectively improve defensive disadvantages. Honeytoken, a type of cyber deception method, can effectively detect data leakage and abuse when used properly. This study utilizes a Natural Language Processing (NLP) algorithm to construct an efficient honeytoken system for securing credentials.The algorithm enriches the semantic features of usernames and generates honeytokens. Experimental results demonstrate the ability to produce honeytokens that closely resemble real usernames in both structure and semantics. Additionally, our analysis reveals that the majority of usernames carry semantic information, validating the efficacy of our approach.