Presentation Attack
Ayrıca şöyle anılır spoofing attack, presentation attack instrument, PAI, anti-spoofing
A presentation attack is any attempt to fool a biometric system, a face scanner, a fingerprint reader, a voice check, by showing it something that isn't a live, genuine characteristic of the real person. The simplest version is holding a printed photo up to a camera; more sophisticated versions include playing a video on a phone screen, wearing a realistic mask, or feeding in an AI-generated ("deepfake") face or voice. The defense against any of these is called presentation attack detection, or PAD.
The field distinguishes attacks by sophistication because the right defense differs: a simple print/replay attack can often be caught by texture and reflection analysis alone, while a good silicone mask or a well-made deepfake needs depth sensing, temporal analysis across multiple frames, or infrared imaging to catch reliably. Independent testing labs grade systems against both tiers under a published standard, so a buyer doesn't have to take a vendor's word for how good their anti-spoof defense really is.
Who actually built this
The standard defining this taxonomy and how to test against it is ISO/IEC 30107, maintained by ISO/IEC JTC 1/SC 37, the international biometrics standards committee. Independent test labs (for example, iBeta in the US) run accredited conformance testing against it. Solidus's own PAD model is built on MiniFASNetV2, an openly published architecture from Minivision AI's Silent-Face-Anti-Spoofing project: the engineering credit for the model itself belongs there, not to Solidus.
Solidus today
The model runs in Solidus's inference service, but Solidus has deliberately not cleared it for a public anti-spoof claim: it hasn't been through the adversarial security review that would need to happen first, and no independent lab has tested it against ISO/IEC 30107-3. This is named directly rather than left implicit, because "unaudited, unclaimed" is a meaningfully different, and more honest, status than "shipped."
See also
Liveness Detection is the broader capability presentation attack detection is one half of. KYC and IDV are where this matters in practice.
Nereden geliyor
Bunu başkası belirtti. Solidus bir uygulamasını yazdı.
"Presentation attack" is the formal term from ISO/IEC 30107-1 for presenting a biometric system with something other than a live, genuine characteristic, a printed photo, a video replay on a screen, a silicone mask, a synthetic deepfake, in order to fool it. ISO/IEC 30107-3 defines how independent test labs grade defenses against these attacks (Level 1 covers simple print/replay attacks; Level 2 covers more sophisticated artifacts like masks). Solidus has not been through any such independent test. The specific defensive model Solidus's pipeline runs, MiniFASNetV2, is an open architecture published by Minivision AI's Silent-Face-Anti-Spoofing project; it's credited here, not claimed as Solidus's own invention.
Bunu nasıl doğrularsınız
Test ağında çalışıyor. Ana ağda değil.
None publicly runnable, and no third-party lab result exists to cite. Said plainly rather than substituting a softer claim.