Imaging Sensor PUF Key Generation Using Statistical Pixel Outliers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for securing imaging sensors, such as cameras or displays, lack reliable mechanisms for generating unique identifiers or keys, and there is a need for systems that can leverage the inherent variability in sensor components for secure authentication and data protection without additional hardware integration.
Innovation Solution
A method involving defining a subset of sensor components, challenging them under uniform conditions, and determining statistical moments of their temporal distributions to identify outlier pixels, which are used to generate a unique and reproducible key for authentication and encryption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If statistical moments of order i are used to identify outlier pixels, then key reliability is improved, but computational complexity increases
Solution Approach 1:
The patent pre-calculates and stores the statistical moments (mean, standard deviation, skewness, kurtosis) of pixel signal distributions during a calibration phase. These pre-computed statistical characteristics are then used to quickly identify outlier pixels without performing complex real-time calculations, thus improving key reliability while reducing computational complexity during actual key generation.
2Reliability
If a larger subset of pixels is analyzed, then key uniqueness is improved, but processing time increases
Solution Approach 1:
The patent replaces exhaustive analysis of all pixels with a statistical sampling approach. By analyzing the distribution characteristics (statistical moments) of pixel signals and identifying outliers based on these distributions, the system can achieve unique key generation by examining a smaller, representative subset of pixels rather than processing the entire pixel array, thus reducing processing time while maintaining key uniqueness.
3Measurement precision
If threshold values are made adaptive based on statistical distributions, then authentication accuracy is improved, but system complexity increases
Solution Approach 1:
The patent uses fixed threshold values derived from the statistical properties of pixel signal distributions (such as multiples of standard deviation from the mean). These thresholds are determined during calibration based on the measured statistical moments and remain constant during authentication operations. This approach improves authentication accuracy by adapting thresholds to the actual sensor characteristics while avoiding the complexity of real-time adaptive threshold calculation.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
There is disclosed a method of handling a sensor, comprising the steps of: defining a subset of sensor components of the sensor; challenging said subset under uniform conditions; receiving output signal values from said subset; for each component of the subset, determining the statistical moment of order i of the temporal distribution of the output signal value of said each sensor component; determining one or more outliers sensor components, said outliers sensor components being components whose ith order statistical moment has a difference with the mean value of the spatial distribution of the chosen moment over the subset superior in absolute value to a threshold, the ith order statistical moment of one sensor component being estimated on the temporal distribution associated to this sensor component. Developments describe in particular the use of imaging sensors, key generation, authentication, helper data files and the handling of videos.