What this collection does not establish.
This library is not a census or a random sample. Its cases illustrate documented techniques and outcomes. Their distribution does not establish attack frequency, the share of breaches caused by social engineering, a trend over time or a typical financial loss.
Public disclosure, language, source accessibility and research choices affect coverage. Countries, sectors and recent events with accessible reporting can be overrepresented. A missing case or missing field is not evidence that an event or harm did not occur.
Some records rely on court documents or direct disclosures; others depend on attributed reporting, research or disputed accounts. Source independence matters: several articles repeating one statement do not provide several independent confirmations. Primary-source labels are not accuracy scores.
Citations and legal proceedings change. A cached source describes what was available at its recorded time. A blocked page or failed fetch does not establish that a citation is dead. A dated conviction, sentence or investigation statement should not be treated as a live docket check.
Many incidents have no quantified loss. Other records report estimates, claims, aggregate amounts, court orders or recoveries. These are not interchangeable. Read each amount with its financial basis and note, and consult the source before combining figures. Missing values or export sentinels must not be treated as evidence that the loss was zero. A statistic calculated only from quantified cases is not representative of all incidents.
Structured fields omit detail. An incident can span years, countries, channels and organizations. The narrative and source record should resolve ambiguities that a sortable year, tag or numeric column cannot. Research/advisory records are distinct from individual incidents. Overlapping tags must not be counted as mutually exclusive shares.
AI systems contributed to this library. Earlier drafts and checks left errors that subsequent reviews found. Independent review, source hashes and schema validation reduce different risks; none proves that every statement is correct. The methodology explains the current process without treating historical review labels as a guarantee.
If a claim matters to a decision or publication, verify it against the cited source. Report errors to hello@diopter.ai.
Social Engineering Examples