This paper studies malicious behavioral obfuscation through the use of a new abstract model for process and kernel interactions based on monoidal categories. In this model, program observations are consid-ered to be finite lists of system call invocations. In a first step, we show how malicious behaviors can be obfuscated by simulating the observa-tions of benign programs. In a second step, we show how to generate such malicious behaviors through a technique called path replaying and we extend the class of captured malwares by using some algorithmic transformations on morphisms graphical representation. In a last step, we show that all the obfuscated versions we obtained can be used to detect well-known malwares in practice.