NLP Applied to Cybersecurity
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.
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.