Sensor pattern noise in JPEG compressed images

Sensor pattern noise has been used as a unique fingerprint for various forensic tasks, including source camera identification. However, quantization noise introduced by JPEG compression may impact sensor pattern noise; the robustness of sensor pattern noise is too weak to identify compressed images. This paper proposes a new method to identify the source of JPEG compressed images based on sensor pattern noise. By introducing quantization noise model, a modified model of sensor pattern noise is made to weaken the effects of quantization noise. Experimental results demonstrate the effectiveness of this approach.