IsolationForest is implemented with a bagging of ExtraTreeRegressor and the author of extrmely randomized trees propose a way to generate random (binary) splits by selecting a random subset of
their possible values.
Can IsolationForest handle with categorical attributes in sklearn？

Ref.
https://orbi.uliege.be/bitstream/2268/9357/1/geurts-mlj-advance.pdf