How we publish
Every full report here passed the same pre-publication checklist. These are the rules a report must meet before release.
We never publish the address, hostname, domain, path or username of a host that our analysis treats as breached, exposed or abused. Hosts get neutral labels plus network type and country. Organisations are named only when the fact is publicly confirmed.
An address appears only when several independent signals or workbench evidence place it in adversary infrastructure. Candidate, dual-use and residential hosts are reported as aggregates, and downloadable data replaces them with salted tokens.
Captured tooling is described by purpose, key strings, size and the SHA-256 recorded at acquisition. Reports carry no working script bodies, exploit offsets or partial secrets.
Each CVE is checked against the CVE record or NVD, and each reference link is opened and dated. Labels from upstream feeds are reported as feed labels, not as attribution.
Staged files are described as staged. Links between populations that the data does not show are labelled as an analytical model, and simulations are never presented as field results.
Each rule is loaded on its own engine, checked against the behaviour it claims to catch, and published on the report page byte for byte as printed in the report.
Each report page lists only verified adversary indicators, with CSV, JSON and STIX 2.1 downloads. Candidate, suspected, dual-use and victim-related values stay out of the public list.