< Back to projects

NLP Applied to Cybersecurity

Master 2 - FYC Group of 3 (with Vincent and Nagib) 2025-2026 Completed

A six-chapter report bridging linguistic theory (semantics, morphosyntax, syntax) with practical cybersecurity applications: regular expressions, Unicode forensics, NLP applied to logs and phishing detection, obfuscation, and Shannon entropy. Two hands-on tools were built to back the theory with working code.

From linguistic theory to security tooling Linguistic theory semantics, morphosyntax Unicode analyzer homoglyph detection (Python/spaCy) Obfuscation demo Rust XOR vs strings Applied to logs, phishing
Six-chapter report moving from linguistic theory to two working tools applied to phishing and log analysis.

Unicode analyzer

unicode_analyzer.py (with a spaCy-based variant) flags suspicious or homoglyph Unicode characters - for example a Cyrillic character standing in for a Latin look-alike in a domain name or a piece of text. Directly useful for anti-phishing and forensic triage: the same class of technique used by domain confusable detection in real-world anti-phishing tooling.

Obfuscation mini-project (Rust)

Two comparable binaries, outil_plaintext.rs and outil_ciphertext.rs, built to demonstrate why string obfuscation (here, simple XOR encryption of embedded strings) defeats naive static analysis such as running strings on a binary. An automated test script and a validation report back up the comparison with reproducible results - the plaintext binary leaks its strings instantly, the obfuscated one does not.

Tech stack

NLP spaCy Python Rust Unicode / homoglyph analysis String obfuscation Shannon entropy