7 solutions for real-time liveness detection and deepfake fraud

What is real-time liveness detection, and how does it help combat deepfake fraud?

Real-time liveness detection checks that a real, physically present person is on the other end of an identity verification session. It’s built to catch someone pretending to be somebody else, whether that’s with a photo, a pre-recorded video, a mask or another spoofing trick. You’ll usually run into it during selfie capture or video-based verification.

To tell a live person from a fake, the system looks at biometric and visual signals, like facial movement, skin texture, depth cues, lighting or other signs of authenticity. While the session runs, it usually returns a result you can act on, such as approve the applicant, reject the attempt or ask for additional verification.

There are two main approaches:

• Active liveness detection asks users to do something, like blink, turn their head or follow instructions on the screen. Those challenges show the person is engaging in real time, but they add friction.
• Passive liveness detection works in the background instead. It analyzes signals without asking the user to do anything extra, so verification is typically faster and less disruptive.

So, how does this help with deepfakes? Fraudsters can use AI-generated images, face swaps, or manipulated videos to pose as genuine customers during onboarding or authentication. Liveness detection helps flag attempts to pass off synthetic or manipulated faces as a real person, especially when the fake is held up to a camera.

That said, solutions vary in how well they protect against convincing deepfake attacks. For example, some are designed mainly to catch physical presentation attacks, like a printed photo, while others add specialized capabilities for detecting AI-generated media and digital injection attacks.

Why liveness detection isn’t enough: you need a layered approach

Liveness detection answers if biometric authentication evidence comes from a genuine, physically present person. However, passing a liveness check doesn’t make someone genuine.

How so? Well, a fraudster can be a real, live person committing identity theft using a stolen identity, a manipulated document, or a synthetic identity. And sophisticated attackers may skip the camera altogether by injecting manipulated media directly into the verification process.

That’s why a layered approach works best. Pairing liveness detection with other verification methods lets you assess the biometric capture, the identity evidence, and the person’s broader risk profile together. Here’s what to look for:

• Document verification: Checks identity documents for signs of forgery, manipulation or tampering to flag when a real person tries to verify themselves with fraudulent evidence.
• Identity data verification: Cross-references submitted personal information against trusted data sources to see whether the claimed identity exists and the details line up. It protects against stolen or fabricated identities that could otherwise pass biometric checks.
• Device and behavioral risk signals: Look at suspicious device activity, unusual application patterns, or repeated verification attempts. They can surface coordinated fraud that wouldn’t be obvious from a single selfie or document.
• Cross-industry fraud intelligence: Draws on identity data from across many sectors, so you can spot suspicious patterns that don’t show up in your own customer data.
• Risk-based orchestration: Brings all the results together and decides when to look harder. An applicant who passes liveness and document checks with no other risk indicators can move ahead, while one with inconsistent identity data or suspicious device signals may need further verification.

Mitek Systems

Mitek¹ specializes in digital identity verification and mobile image processing using AI. Its MiVIP platform can layer data, documents, biometrics, risk insights and AML screening, so liveness is one piece of a bigger workflow.

Mitek’s IDLive Face passively detects fraud from a single image, with no blinking, smiling or head turning required. It picks up natural face movements when they’re present and detects injection attacks from virtual or external cameras. Because it runs quietly in the background, it can be harder for fraudsters to tell it’s there and adapt their approach. Mitek reports testing the technology for consistent performance across race, gender and age.

IDLive Doc looks for uploaded, injected or replayed documents to tell genuine ones apart from digital reproductions in real time. It also targets high-resolution screen images meant to slip past verification, altered portrait photos on IDs and overlays, swaps and other signs of physical forgery.

Microblink

Microblink⁴ treats liveness detection as one signal among many, alongside AI-powered document, biometric and fraud checks, in its Identity Intelligence OS platform. Each check feeds into risk scoring, adaptive workflows and ongoing evaluation of the person behind the account.

Its passive face check analyzes depth cues, surface texture and lighting behavior to distinguish genuine physical presence from spoofed, synthetic or replayed inputs. Autocapture and on-device feedback help make the experience smoother for genuine users.

Microblink also offers liveness for documents and cards:

• ID document liveness confirms a physical document was in front of the camera, not a reproduction or digital proxy
• Screen detection flags documents displayed on a mobile or desktop screen
• Photocopy detection evaluates whether a submission is a color or black-and-white reproduction
• Payment card liveness confirms a real card was physically present during image capture

A further layer analyzes each document for generative AI manipulation, deepfake artifacts and synthetic identity signals.

Jumio

Jumio⁵ is an AI-powered identity verification platform, with liveness detection that uses a patented active illumination technique. It’s designed to give enhanced protection against deepfakes, injection attacks and presentation attacks. Automated capture gives users real-time feedback so they can take a good selfie with minimal friction.

The selfie technology checks that the user is real (rather than a photo, video or paper copy), isn’t wearing a mask and is physically present at the moment of detection. Jumio can also combine biometrics and liveness with other signals in one platform for a layered defense.

iProov

iProov⁶ is a biometric identity security provider, and it offers two liveness options:

• Express Liveness is its low-friction option. Multi-frame liveness detection analyzes the image for spoofs presented to the camera, like 2D and 3D masks, photos and presented videos. It also checks the device for spoofs injected into the system, like emulators, virtual cameras and man-in-the-middle attacks.
• Dynamic Liveness is a higher-security, passive challenge-response option. Its patented Flashmark technology sends a unique color code to the device, and the reflection of light helps confirm the person is authenticating in real time. iProov positions it as advanced protection against AI-generated attacks like deepfakes.

iProov’s face authentication also uses active bias mitigation to help keep performance consistent across age, gender and skin tone.

Strengthen your defenses against evolving fraud methods. Book a demo to learn how we can help.