Adaptive Policy Management for Network Security and Human-Automation Teamwork

Advanced computing has increased the potential for applications to work with less human oversight. They can perform tasks that would be impractical or impossible using traditional software. However, this additional capability, if unchecked, also has the potential of effecting severe damage to operations, if the code is buggy or malicious.

Policy management protects networked systems against specific forms of misuse, such as unauthorized access to information. IHMC researchers have pioneered techniques to apply sophisticated policies to all aspects of system behavior. These techniques use semantically rich policies to ensure that complex networked systems and intelligent automation operate under established guidelines and will be continually responsive to human control.

IHMC’s KAoS policy services framework is the most flexible and powerful policy management system available. It uses the OWL (Web Ontology Language) web standard to represent and analyze policies for consistency, then “compiles” and distributes them on the fly in an extremely efficient form for execution. Because of its speed and responsiveness to dynamically changing conditions, KAoS is in use in dozens of demanding projects sponsored by government organizations such as DARPA, NASA, ONR, AFRL, and ARL, as well as large commercial firms.

In addition, KAoS works in conjuction with several other IHMC research thrusts. In network and security research, KAoS helps to streamline the flow of sensitive information to soldiers and intelligence analysts, while making the most of limited connectivity and computing resources. By integrating KAoS policies and offloading many low-level coordination tasks to automated components, these systems allow human attention to remain focused on critical tasks, avoiding costly errors. Probabilistic policy learning and adaptation mechanisms can deal with critical uncertainties and risks. KAoS can possibly reduce dependence on cumbersome and labor-intensive methods, which rely on scarce, specially-trained personnel.

KAoS also is integrated into IHMC’s advanced human-computer interaction efforts (e.g., augmented cognition, sensory substitution, speech- and gesture-based interaction) for predicting adversarial behavior. In addition, KAoS contributes to human-automation teamwork. For software agents and robots to participate in teamwork alongside people carrying out complex real-world tasks, they must emulate the qualities that enable natural and effective human teamwork. Developers of such systems also need tools and methodologies to assure reliable and safe integration, even when they are designed independently and operated with reduced human oversight. Both of these goals can be addressed through the use of policy management, informed by ongoing studies of how humans succeed and fail in coordinating joint activity.