Degree of difference sorts data

Researchers from the Institute of Applied Physics in Spain have found an efficient way to sort inconsistent sets of data into groups that share some similarity.

Grouping objects that are similar in some way is the first step in many types of data searching and analysis.

Being able to group sets of data that do not necessarily share the same characteristics, such as disparate Internet attacks or incomplete telephone surveys, is more difficult than grouping data whose pieces match up.

Key to the researchers' method was comparing the edit distance, or the minimum number of elementary edit operations -- like deletions, insertions, substitutions -- needed to transform one piece of data into another. The method allowed the researchers to sort the data without using a hierarchical structure.

They used the method to classify different types of Web attacks, grouping them by severity. This more fine-grained way of classifying Web attacks should help programmers build more secure software, including better intrusion detection systems and firewalls, according to the researchers.

The method is ready to be used in practical applications, according to the researchers.