The Curious Engineer

All blogs / Security

Damien Desfontaines' personal page (outdated version)

desfontain.es

SecurityMath

’17’18’19’20’21’22’23’24’25’26
Posts
70 from 16 Jul 2017 to 25 Oct 2026
Posting
monthly (18 in the last year)
Kind
personal blog

2026 15

  1. Fun privacy bugs: Type confusion in bindings
  2. Fun privacy bugs: Numeric conversion goes wrong
  3. Conference highlight: EuroTDP 2026
  4. A noise calculator for differentially private queries
  5. How and why Google will share anonymized search data with their competitors
  6. Observations on LLMs for privacy work
  7. One year of Hiding Nemo
  8. Banning noise will be a disaster for statistical data products
  9. Guest post: Goodbye (ε,δ), hello μ! Reporting privacy guarantees in machine learning
  10. Notes from Dagstuhl: biomedical data sharing
  11. Research highlight: Cliopatra: Extracting Private Information from LLM Insights
  12. Research post: Privacy in theory, bugs in practice
  13. Guest post: Balancing your dataset? Mind the privacy leaks!
  14. Three kinds of “privacy auditing”
  15. On “simplifying” GDPR's definition of personal data

2025 12

  1. Favorite reads of 2025
  2. Mapping ISO standards to modern privacy attacks
  3. Tell me what to write
  4. Research highlights: Privacy attacks on statistical data
  5. Five hard lessons learned by privacy engineers
  6. Answering the BfDI's questions on personal data in LLMs
  7. Introducing Hiding Nemo 🪸
  8. Can you beat me at WORDPL?
  9. The search for better empirical privacy metrics
  10. A glossary of differential privacy terms
  11. Don't use diffprivlib
  12. Five things privacy experts know about AI

2024 7

  1. Winning at DP VISION — suspiciously easily…
  2. Paper highlights: Utility-boosting geometric tricks
  3. What's up with all these large privacy budgets?
  4. Empirical privacy metrics: the bad, the ugly… and the good, maybe?
  5. Converters between differential privacy variants
  6. Paper highlight: Evaluations of Machine Learning Privacy Defenses are Misleading
  7. Five stages of accepting provably robust anonymization

2023 3

  1. Choosing things privately with the exponential mechanism
  2. Mapping privacy-enhancing technologies to your use cases
  3. What anonymization techniques can you trust?

2022 4

  1. Is differential privacy the right fit for your problem?
  2. Research post: Differential privacy under partial knowledge
  3. A bottom-up approach to making differential privacy ubiquitous
  4. Averaging risk: Rényi DP & zero-concentrated DP

2021 8

  1. A list of real-world uses of differential privacy
  2. A friendly, non-technical introduction to differential privacy
  3. Joining Tumult Labs
  4. Don't worry, your data's noisy
  5. Getting more useful results with differential privacy
  6. Demystifying the US Census Bureau's reconstruction attack
  7. Why not differential privacy?
  8. Converting my PhD thesis into HTML

2020 2

  1. The magic of Gaussian noise
  2. The privacy loss random variable

2019 5

  1. A reading list on differential privacy
  2. « What does a privacy engineer do, anyway? »
  3. Local vs. central differential privacy
  4. Research post: Cardinality Estimators do not Preserve Privacy
  5. Almost differential privacy

2018 8

  1. Personal open access policy
  2. Differential privacy in practice (easy version)
  3. « So, how does your part-time PhD arrangement actually work? »
  4. Differential privacy in (a bit) more detail
  5. Why differential privacy is awesome
  6. δ-presence, for when being in the dataset is sensitive
  7. l-diversity, because reidentification doesn't tell the whole story
  8. Book review: Crash Override

2017 5

  1. k-map, the weird cousin of k-anonymity
  2. Book review: Twitter and Tear Gas
  3. Biometrics: authentication or identification?
  4. k-anonymity, the parent of all privacy definitions
  5. Beginnings

Undated 1

  1. From (ε,δ) to attack risk: Hypothesis testing and f-DP