Skip to main content
Loading navigation
Back to challenges

Build a Deepfake Identity Defense Detector with Vercel AI SDK

Generative AI deepfakes present an existential threat to biometric remote onboarding systems. Construct a multimodal deepfake verification backend using Vercel AI SDK that assesses video frames and facial authenticity in real-time.

CybersecurityHosted by Vera
Sign in free to start

What you are building

Develop a TypeScript deepfake detection agent using Vercel AI SDK to analyze video frame artifacts and liveness proofs.

How work is evaluated

Evaluation

Evaluates synthetic deepfake detection precision and false positive rate.

Datasets

Public datasets, plus the private datasets you have access to.

How submissions are scored

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

test_deepfake_flagging

Verify high blur and missing blink flags deepfake

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

f1_score

F1 score on deepfake detection test set • target: 0.96 • range: 0.88-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

  • Use Vercel AI SDK `generateObject` for structured frame-by-frame deepfake feature extraction.

  • Evaluate key facial consistency indicators (blinking rate, edge artifacts, lighting anomalies).

  • Achieve high precision on flagging synthetic deepfake onboarding attempts.

Resources

Dataset notes

Synthetic frame feature measurements extracted from real vs deepfake video streams.

Challenge setup

Evaluates synthetic deepfake detection precision and false positive rate.

Challenge input

JSON containing frame metadata, artifact_scores, and micro_expression_delta

AI SDK

Framework requirement

Evaluated output

JSON with is_deepfake, confidence, detected_anomalies

Evaluation checks
  • Verify high blur and missing blink flags deepfake
  • F1 score on deepfake detection test set • target: 0.96 • range: 0.88-1
Configured evaluation evidence
  • Benchmark: Deepfake Detection Benchmark v1
  • F1 Score target: 0.96
  • 1 public reference case
Configured runtime
  • JavaScript execution harness
  • javascript sandbox (unavailable on Versalist)

Shows the first task, up to three tools, and up to three evidence items.

View technical recipe

Configured tools

Action Space
  • Vercel AI SDK · Required
Orchestration
  • Vercel AI SDK · Required

Evaluation contract

  • test_deepfake_flagging · Weight 1
  • f1_score · Weight 1

Metrics

  • f1_score · Target 0.96

Benchmarks

  • Deepfake Detection Benchmark v1
  • security

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing frame metadata, artifact_scores, and micro_expression_delta

    Output: JSON with is_deepfake, confidence, detected_anomalies

    Execution harness: JavaScript

Evaluation tools

  • AI SDK · agent_framework

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Build a Deepfake Identity Defense Detector with Vercel AI SDK