# Determining Image Sensor Temperature Using Dark Current

## TL;DR

This paper explores the use of dark current response (DSN) for camera fingerprinting and temperature estimation, proposing a more computationally efficient filtering method for forensic triage.

## Contribution

It demonstrates that DSN can be used for camera identification and temperature estimation, and introduces discrete cosine transformation filtering as a faster alternative to wavelet filtering.

## Key findings

- DSN is a viable fingerprinting method.
- Discrete cosine transformation filtering reduces computation time.
- Temperature information can be extracted from DSN.

## Abstract

The state of the art method for fingerprinting digital cameras focuses on the non-uniform output of an array of photodiodes due to the distinct construction of the PN junction when excited by photons. This photo-response non-uniformity (PRNU) noise has shown to be effective but ignores knowledge of image sensor output under equilibrium states without excitation (dark current). The dark current response (DSN) traditionally has been deemed unsuitable as a source of fingerprinting as it is unstable across multiple variables including exposure time and temperature. As such it is currently ignored even though studies have shown it to be a viable method similar to that of PRNU. We hypothesise that DSN is not only a viable method for forensic identification but, through proper analysis of the thermal component, can lead to insights regarding the specific temperature at which an individual image under test was taken. We also show that digital filtering based on the discrete cosine transformation, rather than the state-of-the-art wavelet filtering, there is significant computational gain albeit with some performance degradation. This approach is beneficial for triage purposes.