PRNU-based image processing is a key asset in digital multimedia forensics.\nIt allows for reliable device identification and effective detection and\nlocalization of image forgeries, in very general conditions. However,\nperformance impairs significantly in challenging conditions involving low\nquality and quantity of data. These include working on compressed and cropped\nimages, or estimating the camera PRNU pattern based on only a few images. To\nboost the performance of PRNU-based analyses in such conditions we propose to\nleverage the image noiseprint, a recently proposed camera-model fingerprint\nthat has proved effective for several forensic tasks. Numerical experiments on\ndatasets widely used for source identification prove that the proposed method\nensures a significant performance improvement in a wide range of challenging\nsituations.\n