Neural networks in the brain are sparsely connected, composed of components with an over 50% failure rate, and still amazingly consistent in their high-level behavior over time. We are building models of biologically plausible neural networks to help explain how the brain can protect against a malicious adversary while keeping networks tiny, low power, and easily trained. Using parameters taken from the somatosensory cortex, we have built a simulator to show the relationships between connectivity and severity of possible attacks. Some prior knowledge of distributed system design is helpful, and we'll teach you all of the neuroscience you need to know.