Advanced Face Liveness Upgrade
Budget: $30 – $250 USD
I have a working ASP.NET (VB.NET) + SQL face-recognition attendance portal that already checks for basic liveness by asking the user to say “Hi” while their mouth movement is tracked. The weakness is that a pre-recorded video of the real user passing that prompt can still fool the system.
I now want to harden the liveness layer with a random-phrase verification challenge. On each login the application should show or speak a short, unpredictable phrase, capture the user’s response in real time, and confirm—through synchronized audio–video analysis—that the words were truly spoken live by the person whose face is being matched. No stored photo, looping clip, or deepfake should get through.
Key goals
• Generate and render fresh phrases every attempt, drawing from a configurable list.
• Detect accurate lip-sync and timing, not just audio similarity.
• Hook directly into my existing VB.NET codebase and SQL user tables with minimal architectural change.
• Return a clear pass/fail flag so the current attendance logic remains untouched.
Acceptance criteria
1. Demonstrable defeat of static images and replayed videos during testing.
2. Source code and comments in VB.NET, plus any helper libraries (e.g., OpenCV, MediaPipe, Azure Cognitive Services) properly referenced.
3. Simple setup guide so I can build, deploy, and tweak the phrase list myself.
If you have ideas for extra safeguards—like random head-movements or an optional MFA step—I’m open to discussing them once the core random-phrase check is solid.
I now want to harden the liveness layer with a random-phrase verification challenge. On each login the application should show or speak a short, unpredictable phrase, capture the user’s response in real time, and confirm—through synchronized audio–video analysis—that the words were truly spoken live by the person whose face is being matched. No stored photo, looping clip, or deepfake should get through.
Key goals
• Generate and render fresh phrases every attempt, drawing from a configurable list.
• Detect accurate lip-sync and timing, not just audio similarity.
• Hook directly into my existing VB.NET codebase and SQL user tables with minimal architectural change.
• Return a clear pass/fail flag so the current attendance logic remains untouched.
Acceptance criteria
1. Demonstrable defeat of static images and replayed videos during testing.
2. Source code and comments in VB.NET, plus any helper libraries (e.g., OpenCV, MediaPipe, Azure Cognitive Services) properly referenced.
3. Simple setup guide so I can build, deploy, and tweak the phrase list myself.
If you have ideas for extra safeguards—like random head-movements or an optional MFA step—I’m open to discussing them once the core random-phrase check is solid.