Perturbation-based methods falsify the data before publication by introducing an element of error purposely for confidentiality reasons. This error can be inserted in the cell values after the table is created, which means the error is introduced to the output of the data and will, therefore, be referred to as output perturbation, or the error can be inserted in the original data on the microdata level, which is the input of the tables one wants to create; the method with then be referred to as data perturbation – input perturbation being the better but uncommonly used expression. Possible methods are: rounding; random perturbation; [and] disclosure control methods for microstatistics applied to macrostatistics.
Sources:
NIST SP 800-188 from OECD Glossary of Statistical Terms