Forensically Sound Retrieval and Recovery of Images from GPU Memory
Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering, 2015
This paper adopts a method to retrieve graphic data stored in the global memory of an NVIDIA GPU.... more This paper adopts a method to retrieve graphic data stored in the global memory of an NVIDIA GPU. Experimentation shows that a 24-bit TIFF formatted graphic can be retrieved from the GPU in a forensically sound manner. However, like other types of Random Access Memory, acquired data cannot be verified due to the volatile nature of the GPU memory. In this work a Color Pattern Map Test is proposed to reveal the relationship between a graphic and its GPU memory organization. The mapping arrays derived from such testing can be used to visually restore graphics stored in the GPU memory. Described ‘photo tests’ and ‘redo tests’ demonstrate that it is possible to visually restore a graphic from the data stored in GPU memory. While initial results are promising, more work is still needed to determine if such methods of data acquisition within GPU memory can be considered forensically sound.
Uploads
Papers by Marc Rogers