Reservoir Labs, Inc. — Department of Energy SBIR Phase I: 03a

Description

The DOE estimates that the costs of a cyber attack onto its main high performance computing (HPC) facilities could potentially be in the order of billions of US dollars. To protect its infrastructure against this increasing threat, the DOE Labs are acquiring a new breed of cyber security sensors capable of delivering unprecedented levels of network visibility at speeds up to 10Gbps per sensor. However as these sites are architected using large-scale distributed HPC systems at aggregated rates of hundreds of gigabits or even terabits per second, each deployment requires a growing number of sensors to match the site’s scale. To enable the promise of deep network visibility, these sites are in need of a scalable solution to manage a security ecosystem consisting of an increasingly large number of sensors (tens or even hundreds of units per site) in a productive, cost-effective and efficient manner. Not developing this technology will imply either (1) a loss of key network security visibility or (2) a loss of productivity in attempting to manage a large number of security sensors without a proper scaling solution. How the Problem is Being Addressed We propose to architect and develop Dynamic Logical Analytic Domains (D-LAD), a technology composed of a set of scalable distributed algorithms and interfaces that will allow security operators—including the DOE National Labs and enterprise customers—to deploy ecosystems of multiple sensors to protect their large-scale HPC systems. With D-LADs, sensors will become mutually-aware of each other, enabling properties such as self-reconfiguration, auto-defensiveness and self-healing, continuously adapting to the new operational conditions, traffic loads, and cyber threats while providing deep intelligence and situational awareness to scientists and security operators. Phase I Project The Phase I technical objectives are three-fold: Design of the D-LAD architecture. Design of a set of modules to enable advanced reasoning for large scale distributed HPC systems. Development of a proof-of-concept (POC) prototype. Implementation of a simplified POC to demonstrate the distributed analytical capabilities of D-LADs. Feasibility assessment and technical work plan towards full product development. Feasibility assessment of D-LADs as a commercially viable technology applied to monitoring and securing large-scale HPC systems. Commercial Applications and Other Benefits The solution we are proposing will drastically change the nature of cyber security. Our proposal provides a technology that will enable a new breed of deep network visibility tools to help secure large scale HPC facilities. This technology will play a crucial role into protecting the facilities that scientists and engineers across the nation use in programs such as DOE’s ASCR to execute compute-intensive science applications that are tightly connected with real societal problems in areas including business, utility, financial, education, and critical national infrastructure systems.