Security Assessment of Partially Encrypted Visual Data: Using Iris Recognition as Generic Measure

Security Assessment of Selectively Encrypted Visual Data: Iris Recognition on Protected Samples

“…For data protected but still left intelligible to some extent (as it is the case with Windowed encryption), the iris recognition based protection strength prediction is clearly superior to the IQM based one. Note that this contrasts to the results in [10], where clearly better correlation was found for Absolute encryption.…”

“…Obtained results do not depend on the iris recognition scheme when used for partially intelligible data (which again is different from [10]). For data beyond the recognition threshold with nothing left to be recognised, results do strongly vary, in particular with respect to employed recognition schemes.…”

“…Thus, in this work, we consider encrypted iris sample data and apply the entire iris recognition toolchain to these data in order to investigate if resulting iris recognition accuracy can properly be related to the assumed protection strength of the underlying visual encryption technique. In particular, we will compare the results to our findings when encrypting normalised iris texture [10], which exhibited varying prediction quality depending on the chosen encryption approach.…”

“…First, we conduct an ROC analysis in verification mode and present EER and obviously, higher EER corresponds to better data protection. Second, we perform recognition in identification mode, and provide cumulative match characteristics (CMC) in slightly different manner than usual [10].…”

“…In recent work [10], we have investigated if iris recognition conducted on encrypted visual iris data is suited to assess the security of the used encryption techniques. As visual data to be protected, we have chosen the "normalised iris texture" as obtained by the rubbersheet transform [11] from the iris sample data.…”