Penetration testers and red teamers often need to bypass endpoint and network protections such as anti-virus and IPS to further access into a client’s network and better emulate the behavior of malicious actors. For this purpose, attackers use a number of methods to bypass antivirus solutions and avoid detection, including obfuscating payloads. This talk will discuss a novel technique of hiding payloads inside of neural networks. An open source proof-of-concept encoder and loader called “ENNEoS” (Evolutionary Neural Network Encoder of Shenanigans) will be demonstrated. The encoder uses genetic algorithms to evolve complex neural networks that output the payload shellcode on demand. A high level overview of the technique will be covered, with a more detailed explanation given to how to use the technique.