Guarding Against Malicious Biased Threats (GAMBiT) Experiment 2

The GAMBiT (Guarding Against Malicious Biased Threats) Experiment 2 dataset offers a rich collection of behavioral and technical data from red team cyberattack scenarios aimed at advancing research in attacker modeling and cognitive analysis. This study engaged 20 skilled participants who performed self-paced cyberattacks over two days within a simulated enterprise network. Built on the SimSpace Cyber Force Platform, the environment replicated real-world enterprise infrastructure with approximately 40 virtual machines, routers, switches, and simulated user traffic. Participants began with limited access and progressed through the network by identifying targets, deploying techniques, and exfiltrating sensitive data. To capture attacker cognition and reasoning, the experiment incorporated psychometric evaluations, cognitive reflection tests, and periodic surveys tracking reasoning, mood, and decision-making processes. Operational data such as keylogs, terminal histories, and network traffic were collected and post-processed into derivative datasets. Comparing with Experiment 3, Experiment 2 serves as the control group data where there are no embedded \u201ctriggers\u201d within the environment, such as deceptive cues designed to elicit cognitive biases.