02 / Research
Two questions shape our work: where does the randomness that produces your keys come from, and what does your hardware wallet actually emit? Around them, we study protocols, signing interfaces and supply chains — with the same demand for evidence.
Areas of work
Each area combines documentary review, experimentation and analysis of real-world use.
Reentrancy, access control, oracles, upgrade mechanisms, cross-chain bridges: we analyse the invariants that hold a system together, and the assumptions that bring it down.
From initial entropy to recovery after loss: generation, derivation, storage, backup, signing and succession. We test models — hardware, multisig, MPC, social recovery — against the real threats they face, and we test the quality of the randomness that produces their keys.
The moment the user authorises is the most dangerous. EIP-712, abstract accounts, blind signing, persistent approvals, transaction simulation: we study the gap between what is displayed and what is executed.
Zero-knowledge proofs (ZK), fully homomorphic encryption, post-quantum signatures, aggregation: we follow the promises, the trade-offs and the migration paths. A new primitive often moves risk rather than removing it.
Measuring incidents, tracking fixes, comparing practices and contributing to standards. We bring technical evidence to debates on the MiCA regulation, the DORA regulation and the NIS2 directive, without speaking for any commercial actor.
2027 programmes
Areas of work, not results. Their timeline and deliverables will be specified after the declaration with the préfecture.
Boot entropy, firmware RNG, seed generation, signing nonces and BIP-32/39 derivation. Statistical tests, code review and reproduction on devices we own or are authorised to test.
Target deliverable · Test report + reproduction scenarios
LTE/eSIM, Bluetooth, electromagnetic emissions while a PIN is entered or a seed is displayed, and supply-chain implants. We distinguish a real leak, an artefact and a rumour.
Target deliverable · RF map + exfiltration scenarios
Hardware, multisig, MPC, social recovery: a comparative reading of guarantees, dependencies and failure modes, from individuals to institutional treasuries.
Target deliverable · Comparative note
Priorities: RNG-SCOPE and RF-WALLET structure 2027. KEY-LIFECYCLE provides the common framework on key custody.
Fieldwork: reception and analysis of beacon and time-signal stations from several countries, most recently NTSC at the Chinese Academy of Sciences (CAS), with complete reports. Independent work, with no partnership or implicit endorsement by the organisations cited.
Latest audit
The latest report has its own page. The audit archive remains separate from research.
24 October 2026, 18:00 · Critical alert
Legacy macOS versions generate recovery phrases with critically weak entropy. Affected wallets must be treated as compromised: BTC wallets were emptied by a third party and a KMD balance remains exposed.
Method
The same discipline applies to a research note, an audit or a disclosure.
Objective, perimeter, versions, rules of engagement and exit criteria, written before starting.
Assets, actors, trust boundaries and attack scenarios. What is not modelled remains invisible.
Code review, directed tests, fuzzing, dynamic analysis and user interviews.
Reproduction, CVSS v4.0 scoring, technical evidence and prioritised recommendations.
Coordinated disclosure, publication and verification of fixes. No report left without follow-up.
Policy
Have you identified a vulnerability in a wallet, a protocol or a tool we study? Write to us. We always put user protection before publication, reputation or spectacle.
We confirm receipt and open a private channel with you.
We qualify the report, attempt to reproduce it and identify the people who should be notified.
Target deadline before publication. It may be extended if a credible fix is under way, with written justification.
We publish with the project concerned, or at the end of the deadline, without exposing users to immediate risk.
Publications
Short, dated, reviewed and sourced texts. A single format for technical notes, field analyses and standards proposals.
We prefer to announce finished work rather than a table of contents. The first issue will appear after the declaration with the préfecture; it will look at the wallet attack surface and the gaps between interface and transaction.
Our acknowledged limit: we do not sell exploits, do not publish before a fix and do not accept embargoes that would protect a reputation rather than people. Security is not a sales argument.
See our approach to audit