THIS IS A CONTINUATION OF N00014-14-1-0462 The Robust Software Modeling Tool (RSMT)
Abstract
Statement Of WorkThis proposed research project builds upon and extends their prior work on ?Progressive Model Generation forAdaptive Resilient System Software,? and to create and validate theories, methods, and tools for constructing models of the expected correct execution of cyber-systems and then automatically compare system execution at runtime with these pre-established models to detect deviations that are indicative of cyber-attacks or implementation bugs.ObjectiveThe objective for this research is to design a proof-of-concept for a novel technology that facilitates the development and enforcement of incremental software models produced through synergistic manual and machine learning approaches. The RSMT will not only enable the generation of these models, it includes a self-adaptive runtime component useful for guaranteeing that these models reflect the reality of a running software system.ApproachThe proposed research will use both manually-specified and machine-learned models to formalize the correctexecution of a system. Manually-specified models can provide greater accuracy due to human understanding of the abstractions and intent of a system s architecture, design, and implementation. Manually-specified system models, however, can also suffer from bugs or inconsistency in formulation, due to gaps in the modelers understanding of the system, inability to predict design side-effects, or human error. In contrast, machine-learned models are often poor representations of architectural abstractions, but have high accuracy with respect to the complex runtime characteristics of a system that they have been trained to identify. Each system execution is considered a set of training data for machine learning algorithms. The research challenge, therefore, is that the training data may be incomplete or not fully exercise the system and all intended operational modes.Merit/ONR RelevanceThis research project directly support autonomic computing research program at ONR. The model generated by this research project can serve as a reference essential for enabling system reasoning, a key component for the autonomic computing system.Large and complex systems of software, such as the ones used by the Navy, are difficult to completely verify andsecure. These systems are vulnerable to compromises which take advantage of the architecture, protocol andimplementation weaknesses and flaws. As breaches and compromises have become a fact of computing life, it isimportant that our computing systems can adapt and operate effectively under such conditions. There is a need for an autonomic computing system which can continuously assess its own state/health, capabilities and limitations, and adapt to the situation, at cyber speed, toward maximizing the potential success of Navy missions.Progress Statement??? re-architected the RSMT agent from the ground-up with an emphasis on extensibility, extensibility, and performance.??? created a new class transformation system that is much quicker and handles a variety of corner cases and older class versions and that were unsupported in the Phase I SBIR transformation system.??? created a new runtime trace API that permits more rapid analysis of running code.??? created a dynamic probing mechanism that enables specifications of code to be probed while ignoring others. Probes can be enabled/disabled at runtime as desired. This is in contrast to the shotgun approach utilized during the Phase I where all code was monitored all the time.??? investigated complex issues related to recursion and multi-threading.??? created a Java object backing a call graph model of software behavior.??? created a Java object backing a call tree model of software behavior.??? characterized the performance overhead of utilizing RSMT.??? tested RSMT on open-source Apache Commons software.??? created a test suite demonstrating the deployment of several Java exploits pulled from the National VulnerabilityDatabase (NVD).??? created a d
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 26, 2018
- Source ID
- N000141612148
Entities
People
- Douglas Schmidt
Organizations
- Office of Naval Research
- United States Navy
- Vanderbilt University