I'm working on a anonymization project and I got interested in linking attacks. For simplicity I only look at data in table format, such as xlxx or csv data. To anonymise such data the most common technique is generalization. There are others like synthetic data, changing data, deleting data, etc.. To evaluate the results one can use definitions like k-anonymity, l-diversity or t-closeness.
So far I know, that linking attacks combine the knowledge of two or more tables. Say one knows, that there is an entry for a person in all tables. Then you extract the equivalence classes, to which the already know data about the person leads. Then one can interpret the resulting data to gain more knowledge.
My question: Is there a way to describe this in a more formal way? Are there (good) paper, that describe these attacks? Is there common software, that proceeds this attacks?