Our research > Machine learning and data analytics
Machine learning and data analytics
Breakthrough artificial intelligence / machine learning capabilities
Our researchers create and implement innovative artificial intelligence (AI) and machine learning (ML) technologies to address critical challenges in big data analytics, wireless networking, electronic warfare and cybersecurity.
The key to our efforts is world-class research in developing advanced machine learning capabilities and applying them to problems from network cyber defense and automated target recognition to spectrum analysis, vision, and modeling and simulation. As one example, our breakthrough ML paradigm, Learning Using Privileged Information (LUPI), mimics the process of human learning to greatly enhance the accuracy and speed of ML for data analytics, including prediction and anomaly detection.
Growing use of AI/ML techniques also opens up a new attack surface that sophisticated adversaries can exploit via manipulation of inputs or training data to solicit incorrect decisions from AI algorithms as well as inference of sensitive data used to train ML models by observing their decisions. We are pioneering capabilities to assess AI technologies, detect adversarial attacks, and mitigate them using a suite of protection techniques.
The key to our efforts is world-class research in developing advanced machine learning capabilities and applying them to problems from network cyber defense and automated target recognition to spectrum analysis, vision, and modeling and simulation. As one example, our breakthrough ML paradigm, Learning Using Privileged Information (LUPI), mimics the process of human learning to greatly enhance the accuracy and speed of ML for data analytics, including prediction and anomaly detection.
Growing use of AI/ML techniques also opens up a new attack surface that sophisticated adversaries can exploit via manipulation of inputs or training data to solicit incorrect decisions from AI algorithms as well as inference of sensitive data used to train ML models by observing their decisions. We are pioneering capabilities to assess AI technologies, detect adversarial attacks, and mitigate them using a suite of protection techniques.
Anchor Element Copy for linking on the same page:
https://www.peratonlabs.com?data-scroll-to-anchor=ourwork
Copy for linking from an external page:
https://www.peratonlabs.com?data-anchor-link=ourwork
Our workAI assurance
We are developing techniques to assess ML algorithms for vulnerability to adversarial attacks, performed by targeted manipulation of inputs to cause misclassification, by insertion of incorrectly labeled data to force incorrect decisions, or by other means. We’re also creating novel capabilities for increasing the robustness of ML algorithms to such attacks. Machine learning robustness for cybersecurity The growing use of ML for detecting malicious cyber activity poses new threats as sophisticated attackers can disguise their exploits to evade deployed defenses. We are leaders in research on adversarial ML assessment for intrusion detection systems that depend on ML algorithms. Video analytics We are applying our LUPI ML paradigm to the detection of objects and activities in video of varying quality. Using LUPI, we can train models on high-resolution video and use these models to extract information of interest from low-resolution video at run time. Automated target recognition We are leveraging LUPI as well as advanced synthetic data generation capabilities to develop algorithms that automate the process of identifying targets from variable quality imagery. AI for modeling and simulation (M+S) We’re applying AI techniques to complement resource-intensive M+S tools by efficiently characterizing the input space so that scarce computational resources can be focused on detailed M&S of areas that deserve deeper exploration. Analytics and AI for predictive maintenance We're harnessing sensor data, AI and analytics to improve management of engines, planes, vehicles, and platforms. Our predictive maintenance solutions combine data fusion, sophisticated time series analysis, and efficient AI/LM algorithms for real-time predictive maintenance. Text analytics Our researchers are developing capabilities that enable analysts to derive actionable intelligence from large, unknown and disparate sources of textual data. Capabilities include the ability to answer questions and provide supporting evidence. Social media analytics We have developed a platform for monitoring social and news media outlets to track emerging patterns and trends in near real time. Our solution uses advanced natural language processing and ML algorithms and ingests content in multiple languages to generate actionable insights. Tool for data exploration and transformation We have developed Arroyo, a forensic data analysis and transformation tool that scales to transaction volumes involving terabytes of data and billions of records. Arroyo ingests data in diverse formats including unstructured text documents to extract named entities, events and relationships and to sort, classify, compare, transform and store the data. Anchor Element Copy for linking on the same page:
https://www.peratonlabs.com?data-scroll-to-anchor=resources
Copy for linking from an external page:
https://www.peratonlabs.com?data-anchor-link=resources
ResourcesAnchor Element Copy for linking on the same page:
https://www.peratonlabs.com?data-scroll-to-anchor=services
Copy for linking from an external page:
https://www.peratonlabs.com?data-anchor-link=services
Related services
Anchor Element Copy for linking on the same page:
https://www.peratonlabs.com?data-scroll-to-anchor=requestinfo
Copy for linking from an external page:
https://www.peratonlabs.com?data-anchor-link=requestinfo
|
