Track trials, findings, and clinical evidence as it lands.
Search papers, labs, and topics across Lattice.
This paper introduces an Autonomous Cyber Operations (ACO) gym, a framework for training and evaluating AI models for cyber defense by integrating high-fidelity network emulation, realistic user behavior, and scenario generation. The gym enables the creation of labelled synthetic datasets for DNN-based Intrusion Detection Systems (IDSs) and the development of autonomous defense agents. While a DNN classifier achieved near-perfect recall on adversarial flows within the gym, experiments revealed a catastrophic performance collapse due to domain shift, highlighting significant sensitivity to changes in network composition.
DNN-based IDSs trained in realistic cyber environments exhibit near-perfect in-domain detection but suffer catastrophic performance degradation with minor domain shifts.
This paper proposes an Autonomous Cyber Operations (ACO) gym, a unified framework integrating high-fidelity network emulation, realistic user behaviour modelling for benign network traffic and adversarial profiles, and the generation of cybersecurity scenarios through open toolkits. By supporting live and deferred feature extraction at the packet and flow levels, the gym facilitates the creation of labelled synthetic datasets for training Deep Neural Network (DNN)-based Intrusion Detection Systems (IDSs), as well as developing and evaluating modular components for autonomous defence agents through emulated interactions. While evaluation of a DNN classifier within the gym shows high in-domain performance, achieving near-perfect recall on adversarial network flows, our experiments reveal a significant sensitivity to domain shift, showing that even minor changes in network composition, such as variations in traffic protocol distribution, can lead to a catastrophic collapse in model transferability and detection performance.