William S. Evans will@cs.ubc.ca Christopher W. Fraser cwfraser@gmail.com Fei Ma Fei.Ma@microsoft.com Abstract This paper describes the design, implementation, and application of a new algorithm to detect cloned code. It operates on the abstract syntax trees formed by many com- pilers as an intermediate representation. It extends prior work by identifying clones even when arbitrary subtrees have been changed. On a 440,000-line code corpus, 20- 50% of the clones it detected were missed by previous meth- ods. The method also identifies cloning in declarations, so it is somewhat more general than conventional procedural abstraction.