Dimensions overall score 9.0

GitHub Code Pulse

No public code linked for this paper yet.

• A new dataset, Journalist-provided Deepfake Dataset (JDD), consisting of 255 public deepfakes contributed by over 70 journalists since early 2024, was created and analyzed.method 95%
• A synthetic audio dataset (SYN) of dead public figures was generated.method 95%
• A novel Context-based Audio Deepfake Detector (CADD) architecture was proposed.method 95%
• Context and/or transcripts can significantly improve the efficacy of audio deepfake detectors.result 90%
• Performance of baseline audio deepfake detectors improved by 5%-37.58% in F1-score with context and/or transcripts.result 85%
• Performance of baseline audio deepfake detectors improved by 3.77%-42.79% in AUC with context and/or transcripts.result 85%
• Performance of baseline audio deepfake detectors improved by 6.17%-47.83% in EER with context and/or transcripts.result 85%
• CADD is more robust to adversarial evasion strategies, limiting performance degradation to an average of just -0.71% across all experiments.result 90%
• A novel Context-based Audio Deepfake Detector (CADD) architecture was proposed.method 95%
• A Journalist-provided Deepfake Dataset (JDD) was created, containing 255 public deepfakes contributed by over 70 journalists.method 95%
• The proposed CADD architecture utilizes context and/or transcripts to improve audio deepfake detection.method 95%
• Incorporating context and/or transcripts can improve the F1 score of baseline audio deepfake detectors by 5%-37.58%.result 95%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card