Pauline Bourmeau

About

Pauline Bourmeau is an independent security researcher focused on the human layer within cybersecurity systems. Her work blends deep technical capabilities with the humanities — bringing together AI, linguistics, and threat intelligence with cognitive psychology as the central pivot.

She has consulted on multilingual natural language processing, led deep learning and NLP workshops internationally, and created training materials blending STEM with human factors. As founder of DEFCON Group Paris (DC11331) and contributor to the MISP project, she actively advances collaborative cybersecurity practices.

Previously, Pauline worked as a Threat Intelligence Analyst at Corexalys (Chapvision), contributing to the AIL Framework under a European research initiative and conducting OSINT, HUMINT, and SOCINT analysis to profile threats and investigate APTs. Her expertise bridges AI, linguistics, and human cognition, bringing a unique perspective to information security and intelligence challenges.

In the education sector, she served as a tutor at the University of Paris Descartes, and taught French as a foreign language and literacy instruction to war refugees. She teaches what she researches — her courses and workshops grow directly from active fieldwork.

Pauline is also a paraclimber — a medal-winning athlete with a disability at the French Paraclimbing Championships.

Research Focus

• AI & NLP in Security The critical adoption of AI and NLP in security analysis and detection pipelines — applying natural language processing and large language models to threat intelligence.
• The Human Layer How cognition and bias shape threat perception. The sociolinguistics of narrative attacks and online deception — and training critical thinking as a countermeasure.
• Collaborative Defense Improving how cyber threat intelligence gets shared, with a focus on the human factors behind it’t. Contributor to MISP and the AIL Framework; linguist on the CEF European joint threat analysis network (2020-EU-IA-0260).

Current Research

• Social Perspectives in Intelligence Sharing within MISP Communities

  While technical and procedural aspects of threat intelligence sharing are well-studied, its social and cultural dimensions remain underexplored. This research examines how sharing communities actually operate within platforms like MISP — the incentives, the friction, and the human dynamics that determine what gets shared and what doesn’t.

  The methodology combines platform statistics and technical observations with semi-structured interviews across sharing communities, mapping the gap between how tools are designed to be used and how practitioners actually use them.

  Published as a named research project on the MISP Research Projects page

• AI & Cognitive Load Personal Interest

  An informal, ongoing exploration of how AI can ease the cognitive load on analysts and knowledge workers — supporting attention, memory, and decision-making under pressure rather than adding to the noise. Observations are drawn from security professionals and web creators.

Graduate Education

• Criminology
  MSc 2024
  Conservatoire National des Arts et Métiers (CNAM) Thesis on cybersecurity intelligence sharing
• Software Engineering
  École 42 Paris Peer-to-peer learning model
• Interdisciplinary Research
  Center for Interdisciplinary Research and Innovation of Paris Learning Planet Institute
• General Linguistics
  Université Paris-Descartes UFR Sciences Humaines et Sociales
• Philosophy
  Université Paris-Tolbiac

Writing & Essays

• Feb 2026

  AI transforms

  How computer-vision models “see” the world — edge detection, generation, and what falls between.

• Nov 2025

  Inference

  Drawing conclusions from known or assumed facts — what inference means, and what it means for machines.

• Nov 2025

  Are language models intelligent?

  Whether language is necessary for intelligence — across neuroscience, psychology, linguistics, and computer science.

• Aug 2025

  Using critical thinking against online deception

  Fake news, deception countermeasures, and verifying authenticity in an age of frictionless information.

• Nov 2024

  Narrative attacks on the web

  Psychological attacks that exploit existing biases to manipulate public perception of people, organizations, and ideas.

Security Work — Cubessa

• Incident

  Supply-chain attack: malicious PyPI package lightning 2.6.2/2.6.3

  Incident report on a poisoned release in the Python packaging ecosystem.

• CTI

  MISP taxonomy: AI Safety Benchmark

  A threat-intelligence taxonomy for classifying AI safety benchmarks within MISP.

• Offensive

  Pentest & red-team engagements

  Available via Cubessa

Speaking & Workshops

Projects

• ORCRIme

  Online Resources for Crime Reporting and Investigation — a curated list of open resources for investigating criminal activity online.

• CEF Joint Threat Analysis Network

  Linguist contributor to Connecting Europe Facility, European project 2020-EU-IA-0260 — a cross-border joint threat analysis initiative.

• NLP for Coders

  A free, open course spanning beginner to intermediate NLP, updated for Transformers v5. nlpforcoders.com

• Deep Learning for Security Professionals Upcoming

  A one-day hands-on training on the fastai deep learning library, designed for security practitioners. Covering image classification, transfer learning, and practical applications in threat detection. Delivered at Bsides Luxembourg.

• NLP for Beginners Since 2024

  A long-running workshop delivered at international security conferences since 2024, introducing NLP to non-specialists with no prior machine learning background. The material evolves continuously, incorporating new approaches and tools as the field moves.

• Phishing Detection Engine In Progress

  An NLP-based phishing detection system developed as part of SPOT Project, co-financed by the LU-CID initiative, Luxembourg's Ministry of the Economy. Combines linguistic pattern analysis with machine learning classifiers to identify phishing attempts at scale.

Contact

I read everything. I don’t always reply quickly, but I do reply. If you want to link to something here, go ahead.