Robust Defense against LRS Obfuscated DGAsthrough Domain-Specific Noise and Deep LearningAnalysis

The identification of novel variations within Domain Generating Algorithm (DGA) families is crucial for the security of critical infrastructures. Despite existing research, the adaptability of techniques to diverse DGA manifestations, including Linear Recursive Sequence (LRS) modification, remains a major concern. To address this, this research proposes a comprehensive technique incorporating Damerau-Levenshtein Distance to enhance a Feedforward Neural Network (FNN). This technique offers insights into domain-specific noise, enhancing its applicability and robustness against sophisticated cyber adversaries. Systematic evaluation shows the technique's adaptability to diverse forms of noise, including domain-specific variations, promising applications in real-time threat detection and incident response. The selection of Damerau-Levenshtein Distance is motivated by its ability to quantify minimal operations for transforming strings, providing valuable insights into structural similarities among DGA strings. The approach's adaptability to the dynamic nature of DGAs is emphasized, acknowledging its effectiveness depends on dataset characteristics. The research addresses the challenge of supporting multiple dataset families, each having distinct classes, highlighting the difficulties in managing diverse characteristics and nuances. With an accuracy rate of 94 percent, the technique proves its robustness and versatility for cybersecurity practitioners dealing with diverse cyber threats. The method's success in simultaneously supporting multiple dataset families positions it as a valuable asset for securing network infrastructures.