AWS Comprehend unable to detect card number as PII unless it called out as Card or masking.
I was learning AWS Comprehend and trying to explore its capabilities. It is very nice service. I found one of the scenario where i wanted to check PII identification use case in a conversation. I found when we explicitly called out card number or use marking then it is identifying as PII. But when i used this is my details or number it is not considering as PII though it is 16 digits.! We can use it as learning pattern whenever it is 16 digits or like phone number 10 digit it should highlight as PII with some lower confidence level. So data engineer can check back
Enter image description here- 最新
- 最多得票
- 最多評論
Thanks for bringing this to out notice. Can you tell us more on what do you expect the model to predict the number to be ? Without much explicit context on whether a number is credit card number, it is hard for even humans to quantify the correct PII Type. Please do note that the functionalities' objective is to predict correct the PII type instead of simply classifying if it is a PII or not , so unless it doesn't strictly fit under the definition of a certain PII type that we support you won't see it classfied under any PII type.
Not every 16 digit random number would be a credit card number , credit card numbers also follows Luhn's algorithm to be valid number .
