News
#all
#publication
#conference
#m3aawg
#talk
#work
all tags →
- 22 Nov 2026 A New Look, and a Newsletter
The site got a redesign: a simpler homepage, a News section, tag filters, and a newsletter to get new posts by email. - 2 Nov 2026 Dr. Gabriel Loiseau: PhD Defended!
Gabriel Loiseau defended his PhD on privacy- and utility-preserving text anonymization. A look back at his thesis, his papers, and what comes next. - 29 Oct 2026 M3AAWG 68th General Meeting in Paris
At the M3AAWG 68th General Meeting in Paris, Gabriel Loiseau and I gave a training on getting better results from LLMs without fine-tuning and without wasting tokens. - 9 Oct 2026 New Role: AI SIG Leader at M3AAWG
Since early 2026, I lead the Artificial Intelligence Special Interest Group at M3AAWG, the anti-abuse industry's working group. - 30 Jun 2026 My PhD Students Rock: Three New Publications
Three new papers from my PhD students, on text anonymization, privacy assessment, and intent detection in phishing emails. - 16 May 2026 LREC 2026 with Gabriel and Senaid
At LREC 2026 in Palma, Senaid presented DIDECO, our dataset for intent detection in digital communications, and Gabriel presented our work on distilling privacy assessment. - 19 Feb 2026 M3AAWG 66th General Meeting in San Diego
At the M3AAWG 66th General Meeting in San Diego, I gave a hands-on training on how AI agents get exploited, and the defenses that hold up. - 15 Oct 2025 Tau-Eval: A Unified Evaluation Framework for Useful and Private Text Anonymization
Tau-Eval, accepted at the EMNLP 2025 Demo Track: an open-source framework to evaluate text anonymization on both privacy and utility. - 13 Jun 2025 Detection of Cyberthreats with Computer Vision (Part 3)
Part 3 of my Hornetsecurity series on image analysis: content-based near-duplicate detection, such as object recognition and embedded text. - 5 Jun 2025 M3AAWG 64th General Meeting in Vancouver
At the M3AAWG 64th General Meeting in Vancouver, my first one on site, I gave a talk on how LLMs can leak sensitive information, and how to defend against it. - 21 Mar 2025 TAROT: Task-Oriented Authorship Obfuscation Using Policy Optimization Methods
TAROT, accepted at PrivateNLP 2025: rewriting text with small policy-optimized models to hide its author while keeping it useful. - 26 Feb 2025 Detection of Cyberthreats with Computer Vision (Part 2)
Part 2 of my Hornetsecurity series on image analysis: spotting duplicate and near-duplicate phishing images with hashes and color histograms. - 9 Apr 2024 WikiPhish: A Diverse Wikipedia-Based Dataset for Phishing Website Detection
WikiPhish, accepted at CODASPY 2024: a phishing website dataset of 110,606 pages, with Wikipedia as a diverse source of legitimate ones. - 30 Sep 2023 Targeted Attacks: Redefining Spear Phishing and Business Email Compromise
Our arXiv preprint proposes one definition of targeted attacks, covering spear phishing and Business Email Compromise, and reviews how to detect them.