Advancements in artificial intelligence and machine learning (AI/ML) have been a catalyst for explosive change across many industries—but they have also added complexity to the threat landscape by creating new vulnerabilities.
From cutting-edge applied research and development to deployment, Peraton is a leader in harnessing the power of AI/ML—addressing government’s mission-critical challenges while developing best-in-class solutions that are setting new boundaries in the fields of robotics and autonomous systems, electronic warfare, communications, and information analytics, including:
- A platform to integrate modular RF systems
- Intelligent data retrieval for clinical trials and public research
- Multi-faceted detection that stops disruptive Trojans at the source
- A paradigm to accelerate training for machine learning
- New methods to introduce AI into the design process
Challenge:
Leverage machine learning to drive multi-domain operations and enable JADC2
Modern warfare requires coordinated, real-time operations across multiple domains—land, air, sea, space, and cyber. This, in turn, requires multi-function platforms with the ability to orchestrate diverse activities, manage shared resources, and determine optimal actions—while also incorporating situational awareness at the tactical edge and adapting automatically to changes.
Peraton Labs has developed a multifunction radio frequency (RF) platform manager that uses sophisticated machine learning techniques to optimize resource allocations and dynamically adjust to changing objectives and conditions. The platform manager delivers superior performance for real-time resource management on modular RF systems supporting communications, electronic warfare, and radar. By supporting modular open systems approach interface standards, the solution can easily scale to multiplatform battlefield management and orchestration.
Challenge:
Quickly answer biomedical questions when faced with a pandemic
The COVID-19 pandemic—and the surge in clinical studies in response to it—has highlighted the need for smart resources that can help researchers quickly and efficiently find answers to specialized biomedical questions when time is of the essence.
Peraton’s Analytics for Acquiring Clinical Knowledge (ATACK-COVID) system can answer questions in multiple languages, generating sentence-level responses from a dynamic collection of clinical trials and research reports. By using publicly available data enriched with contextual information, this solution can adapt the latest neural language models to produce state-of-the-art accuracy in yes-no question answering. Quick access to high-quality answers from clinical trial research can dramatically reduce the timeline from the identification of therapeutic drugs to the implementation of treatments that reduce the severity of future global pandemics—improving patient survival.
Challenge:
Secure AI/ML systems to prevent life-threatening risks
As reliance on AI/ML increases, so do the risks. Trojans are often hidden within AI systems to cause malicious interference and disrupt operations. A compromised system could have wide-ranging consequences—vehicular crashes, missed diagnoses, and stolen identities—or failures in intelligence, surveillance, and reconnaissance, and financial risk management.
Hackers can attack ML algorithms during the training phase and disrupt AI systems in development causing them to make erroneous decisions, such as misidentifying road signs or failing to detect cyberattacks. Peraton Labs develops solutions to secure and defend AI systems. Trojans in Artificial Intelligence (TrojAI) is one such effort, providing a multi-faceted detection solution that can automatically detect Trojans hidden within AI systems, then assess and mitigate the threats.
Challenge:
Teach machines to learn more like humans
Critical challenges in big data analytics, wireless networking, electronic warfare, and cybersecurity are outpacing the ability and speed of existing machine learning (ML) methods. Learning Using Privileged Information (LUPI) is a ML paradigm that mimics the process of human learning to enhance the accuracy and speed of ML for data analytics, predictions, and anomaly detection.
The use of ML can be severely hampered by limitations in the amount or quality of training data. With LUPI, ML systems take advantage of so-called privileged information for much faster and more accurate learning. Like a human learner, LUPI harnesses information available only during training and leverages it in ML applications. These applications learn more accurately and quickly and require many fewer training examples. LUPI has been successfully applied to a variety of applications such as video analytics, object recognition, and target identification.
Challenge:
Shorten the solution cycle to unlock the power of autonomous systems
The design and control of complex cyber-physical systems (CPS) such as robots, unmanned underwater vehicles, and other autonomous platforms are determined by many parameters and configuration settings. As a result, the solution space for robots and other complex autonomous systems is vast and filled with possibilities.
Peraton’s Autonomous System Design and Control blends AI and optimization methods to produce fast and automatic designs and repairs for CPS. The algorithm is extremely efficient at including more factors in the design space to produce better solutions. This approach factors in the operational environment and mission requirements, and accounts for degradations or alterations in the system. It’s a technique that significantly shortens the solution cycle, greatly reduces effort and cost, guarantees operational accuracy and safety, and enhances mission innovation.
