Profile picture forensics
The avatar is downloaded and measured — image entropy, distinct colour count, edge detail,
skin-tone distribution — to tell a real photograph from a logo, a silhouette or the
platform's default placeholder. Each picture is perceptually fingerprinted, so a photo
reused across a network of fake accounts can be recognised.
Registration date
Platforms do not publish signup dates, but they encode them in account IDs. On X the date
is exact. On TikTok it is close. On Instagram and Facebook it is an era rather than a day —
and the report tells you which of those you are looking at.
Audience arithmetic
Followers gained per day since signup. Followers against accounts followed. Likes against
follower count. Bought audiences survive a glance at a profile; they do not survive the
arithmetic.
Handle and name construction
Script-generated usernames have measurable signatures: consonant runs, vowel ratios and
letter-pair frequencies that human-chosen names do not produce. We also detect
impersonation patterns — digits standing in for letters, authority words with numeric tails.
Bio and personal details
Investment and crypto vocabulary, promises of returns, instructions to continue on
WhatsApp or Telegram, links routed through shorteners, phone numbers — the standard
furniture of a scam approach.
Evidence you can audit
Every finding is listed with what was observed and how much it moved the result. If you
disagree with a signal you can see exactly which one it was, instead of arguing with an
unexplained percentage.