San Diego, Calif. – June 26, 2018 – ObjectSecurity, a leader in solving complex, evolving defense and industrial cybersecurity and supply chain risk challenges, today announced that it has been selected for a Small Business Innovation Research (SBIR) contract from the US Missile Defense Agency (MDA).
As stated in the original solicitation titled “Artificial Intelligence to Simulate Cybersecurity Red Team“, the objective is to “design, develop and demonstrate an innovative artificial intelligence system to simulate a Cybersecurity red team. […] The government is interested in innovative approaches for simulating Cybersecurity attacks to train missile defense network defenders. Traditionally, this is done using either canned scenarios or human-in-the-loop training events. Canned scenarios provide limited realism for training while human-in-the-loop events are cost prohibitive. Turning training into a “game” with rewards would incentivize the user to train more frequently to maintain top skill levels. An innovative approach to solve this problem may be an artificial intelligence system that could adapt its attacks real-time based on the trainee responses. For example, a highly trained neural network that could adapt to the trainee’s input with ever-increasing and adapting attack vectors could be one approach to solve this problem. Using a fuzzy logic rule based system might be another approach. […] Develop and demonstrate a gaming concept for training a user to defend a network against multiple threats of varying types and capabilities. The concept should provide feedback to the game participant in a quantitatively measurable format. It should also provide the capability to compare these “scores” based on the participant’s alternatives or courses of action.”
ObjectSecurity’s proposal “WhizRT: Simulated Intelligent Cybersecurity Red Team“ included “an innovative solution that includes a novel, automated and intelligent/adaptive red team attacker engine, together with a virtualized playing field testbed (with lifecycle automation, adaptive scoring and more). WhizRT will be able to intelligently choose its actions through an attack behavior tree and execute attack steps across the hands-on playing field. It combines state-of-the-art machine learning with attack tree models and automated attack execution. The playing field is an actual networked IT environment hosted within a virtual machine environment. Using artificial intelligence (AI) techniques (esp. artificial neural networks), the attacker can adapt its choices of attack steps over time by learning from the defenders behaviors, successes/failures etc.The project revolves around two broad areas, the WhizRT intelligent red team system (based on AI, attack behavior trees, and attack step execution), and the training game concept and testbed (based on interconnected VMs with lifecycle management, scoring, education etc.). WhizRT will be developed to be flexible, and in particular to be deployable in other cybersecurity training game platforms. We are highly qualified and positioned to make the proposed effort a success, and have proven to provide exceptional quality and value (incl. several SBIRs).”
ObjectSecurity’s CEO Dr. Lang subsequently presented the technology at ToorCon 2021 in San Diego, CA as part of the presentation “AI HACKER! Automatic vulnerability assessment & pen-testing of embedded & other systems“: We presented “the results of our government-funded R&D to develop an intelligent automated “vulnerability assessor and penetration tester (VAPT)”, which “intelligently automates the tasks of an ethical hacker (penetration tester). It automatically executes sequences of reconnaissance and exploit actions via the network, finding systems on the network, discovering vulnerabilities, and exposing them. […] For intelligent AI-driven action selection, the prototype includes an AI agent that learns over time and adapts a bit like a human vulnerability assessor or pen-tester, selecting the most promising sequence of actions.”
The technology, which was using advanced AI/ML approaches such as deep reinforcement learning, was subsequently used in ObjectSecurity’s follow-on research, which was fully commercialized in ObjectSecurity’s BinLens™ (formerly ObjectSecurity OT.AI Platform)
About ObjectSecurity
ObjectSecurity LLC is a leader in solving complex, evolving defense and industrial cybersecurity and supply chain risk challenges that threaten national security and production downtime. Our novel research and development are applied to commercial solutions proactively addressing the core source of cyber vulnerabilities and risk – software code and data. Our holistic, proactive approach prevents cyber attacks and disruptive production downtime across industries that support global citizen communities, including military defense, municipal smart cities, public and private transportation, energy, wastewater treatment, power utilities, manufacturing, and the life sciences. For more than 20 years, ObjectSecurity has been delivering cybersecurity and supply chain risk management solutions, including to U.S. defense and federal government agencies. These advanced technologies are now commercially available for government and private sector use. ObjectSecurity is privately-held with headquarters in San Diego, CA, with global representation. Learn more about ObjectSecurity at https://www.objectsecurity.com.
Further information
- SBIR Solicitation MDA18-001, “Artificial Intelligence to Simulate Cybersecurity Red Team”, https://www.sbir.gov/node/1413627
- SBIR Award Information, “WhizRT: Simulated Intelligent Cybersecurity Red Team”, https://www.sbir.gov/sbirsearch/detail/1590191
- ObjectSecurity’s BinLens™ (formerly ObjectSecurity OT.AI Platform), https://objectsecurity.com/otai
- Dr. Lang, “AI HACKER! Automatic vulnerability assessment & pen-testing of embedded & other systems”, ToorCon 2021, San Diego, CA, 2021, https://talks.toorcon.net/toorcon21/talk/PEMWVR/
- Contact us: https://objectsecurity.com/contact
Watch a brief video of ObjectSecurity’s “AI hacker” deep reinforcement learning R&D in action (continued in Navy SBIR R&D):
