Significance of image brightness levels for PRNU camera identification.

Area of Science:

• Digital image forensics
• Computational imaging
• Pattern recognition

Background:

• Photo response non-uniformity (PRNU) is a unique camera fingerprint used for source identification.
• Existing camera identification algorithms, like the court-approved PRNU method, have established error rates on large datasets.
• These error rates do not account for variations in image brightness, which can be affected by camera exposure settings.

Purpose of the Study:

• To investigate the impact of image brightness on the error rates of a court-approved PRNU-based camera identification algorithm.
• To analyze how different brightness levels (dark, nominal, bright) affect the accuracy of forensic image analysis.
• To propose a more accurate characterization of PRNU algorithm error rates by incorporating image brightness.

Main Methods:

• A novel classification method was used to assign brightness levels to images in a large dataset.
• The court-approved PRNU algorithm was applied to images categorized by brightness.
• Statistical analysis was performed to compare error rates across different brightness levels.

Main Results:

• Statistically significant differences in error rates were observed between images of varying brightness.
• Dark and bright images exhibited different error rates compared to nominal images.
• The accuracy of the PRNU algorithm is demonstrably influenced by the brightness of the acquired image.

Conclusions:

• Image brightness is a critical factor affecting the performance of PRNU-based camera identification.
• The error rates of the PRNU algorithm can be more accurately assessed by considering the brightness of the questioned image.
• This finding has implications for improving the reliability of digital image forensics in legal proceedings.