The rise of the internet of things will further increase the number of connected devices, which will provide a new attack surface for compromising computer networks. Network intrusion detection systems have already been proposed to help solve this problem, but they are typically computationally too expensive to run on embedded or IoT devices. In this research, we investigate how binarized neural networks can be used to monitor network traffic on such devices, and flag malicious behavior. We train a binarized neural network classifier on the UNSW-NB15 data set and achieve an 82.1% accuracy. We implement the binarized neural network in hardware using a generator written in Chisel and integrate it on a Zynq-7000 system-on-a-chip. We show that such a system can be implemented on even low-end FPGA devices and can increase the security of future networks.