Does the "wave your hand in front of your face" deepfake test still work in interviews?
TL;DR: No. Security researchers and recruiters report the "wave your hand" and "hold up three fingers" tricks are being defeated by newer real-time deepfake filters in 2026. These manual checks only ever caught cheap, laggy filters and were never reliable evidence on their own. Hiring teams need continuous, multi-signal verification — ID-to-selfie match, liveness, virtual-camera detection — instead of a single improvised gesture test.
The claim
Ad-hoc gesture challenges (wave your hand, cover your face, turn your head) became a popular 2024-2025 recruiter workaround for spotting live deepfake video filters, because early face-swap tools glitched visibly when a hand passed in front of the face.
The evidence
Multiple 2026 LinkedIn posts and threat-intel commentators tracking DPRK-linked operators now say the trick "no longer works" against current-generation filters, and that the same is true of the three-finger test. One widely shared recruiter account described running a full ad-hoc checklist on a single call — name a street near your apartment, wave your hand, turn your head — with the candidate dodging or refusing each request, which is itself a stronger tell than any single test passing or failing. Separately, DOJ and Microsoft's action dismantling 29 laptop farms (roughly 200 computers) tied to North Korean IT workers, with one facilitator sentenced to 8.5 years over a $17M scheme, shows the operators behind these interviews are well-resourced and iterate quickly on detection evasion.
Comparison table
| Method | What it catches | Why it's failing in 2026 |
|---|---|---|
| Wave hand / 3-finger test | Cheap 2023-era face filters that glitch on occlusion | Modern filters render occlusion correctly |
| Ask candidate to turn head | Filters with a fixed camera angle | Newer models track head pose in real time |
| One improvised question | Nothing structural | Easily rehearsed by a coached candidate |
| Continuous multi-signal verification (ID+selfie match, liveness, virtual-camera detection, trust score) | Structural mismatch across the whole session, not one moment | Attacks all layers simultaneously, so a single trick can't defeat it |
Step-by-step for hiring teams
- Stop relying on any single improvised gesture as proof of a real person.
- Require ID-document capture plus a liveness selfie before the interview starts, not during it.
- Run continuous face verification through the full call, not a one-time check.
- Flag virtual-camera and known deepfake-software signatures automatically rather than eyeballing video quality.
- Escalate any refusal to comply with a verification step — that refusal is a stronger signal than any single visual glitch.
FAQ
Why did the wave-your-hand trick stop working? Because current real-time face-swap and avatar tools render hand occlusion and head turns convincingly, whereas 2023-2024 filters glitched or lagged when something passed in front of the face.
Is there a reliable manual test recruiters can still use? No single manual test is reliable on its own. Asking a candidate to name a nearby street, wave, or turn their head can add friction, but only structured refusal patterns — not passing or failing one test — are meaningful signals. Pair them with software-based verification.
What should replace ad-hoc gesture tests? ID-to-selfie identity verification at check-in, continuous face verification through the session, and virtual-camera/deepfake-signature detection — the layered approach Neuroxa's AI Meeting Proctor runs automatically.
Does this affect exams as well as interviews? Yes. The same filter-evasion trend applies to any remote proctored session, which is why presentation-attack detection is increasingly required for online exams too, not just hiring.
By Pinal Dave Last updated: 2026-08-06