Detection device, detection method, and detection program
A detection device includes processing circuitry configured to acquire data to be detected and normal reference data, calculate a Learned Perceptual Image Patch Similarity (LPIPS) distance between the acquired data and the reference data, and classify the acquired data into either a Clean Sample or an Adversarial Example by using the calculated LPIPS distance.
1 . A detection device comprising:
processing circuitry configured to:
acquire data to be detected;
acquire normal reference data separately from the data to be detected;
calculate a Learned Perceptual Image Patch Similarity (LPIPS) distance between the acquired data and the reference data; and
classify the acquired data into either a Clean Sample or an Adversarial Example by using the calculated LPIPS distance,
wherein the processing circuitry is further configured to determine a minimum value among a plurality of calculated LPIPS distances and to classify the acquired data using the determined minimum value.
2 . The detection device according to claim 1 , wherein the processing circuitry is further configured to classify the acquired data into either a normal Clean Sample or an abnormal Adversarial Example by using a model for classifying the data into normal or abnormal, using the calculated LPIPS distance as an abnormality score.
3 . The detection device according to claim 2 , wherein the processing circuitry is further configured to learn the model for classifying data into normal or abnormal, using the calculated LPIPS distance as an abnormality score.
4 . The detection device according to claim 1 , wherein the processing circuitry is further configured to acquire a sign photographed by a camera, as the data to be detected, and
detect an Adversarial Example.
5 . A detection method to be executed by a detection device, the detection method comprising:
acquiring data to be detected;
acquiring normal reference data separately from the data to be detected;
calculating a Learned Perceptual Image Patch Similarity (LPIPS) distance between the acquired data and the reference data; and
classifying the acquired data into either a Clean Sample or an Adversarial Example by using the calculated LPIPS distance,
wherein a minimum value among a plurality of calculated LPIPS distances is determined and the acquired data is classified using the determined minimum value.
6 . A non-transitory computer-readable recording medium storing therein a detection program that causes a computer to execute a process comprising:
acquiring data to be detected;
acquiring normal reference data separately from the data to be detected;
calculating a Learned Perceptual Image Patch Similarity (LPIPS) distance between the acquired data and the reference data; and
classifying the acquired data into either a Clean Sample or an Adversarial Example by using the calculated LPIPS distance,
wherein a minimum value among a plurality of calculated LPIPS distances is determined and the acquired data is classified using the determined minimum value.
7 . The detection device according to claim 1 , wherein the processing circuitry is further configured to acquire a prediction class of the data to be detected.
8 . The detection device according to claim 1 , wherein the detection device is incorporated into an autonomous vehicle.