Alexis Networks, Inc., https://alexisnetworks.com/, is proud to announce a multi-year R&D collaboration with the New York University Tandon School of Engineering, https://engineering.nyu.edu/, to significantly advance the computational science behind unsupervised anomaly detection.
Anomaly detection is an emerging field in machine intelligence, enabling the detection of anomalous and malicious data points which do not follow an embedded, inherent pattern of healthy clusters of normal data. The nature of anomaly is related to the nature of data, however the mathematical and algorithmic underlying mechanism of anomaly detection methods can be adapted to various applications, for example, the anomaly can be caused by a fault in a system such as a sensor fault in an autonomous robotic system or autonomous vehicle. It can also be caused by communication faults and security issues in multilateral and multiagent systems. Other applications are fraud detection (for example for credit cards), cyber-security. For detecting anomalies, gray box and black box models have been used in the literature some of which rely on probabilistic distributions of the data and some rely on neural network models. Despite the great success and advancements in the field, the anomaly detection algorithms still suffer from several factors, including (a) the sensitivity to the amount of available labeled data, (b) dynamics and variation in the normal behavior of the data, (c) noise in data, (d) similarity between advanced malicious activities/events and normal data. This R&D project that will be led by Prof. Atashzar at NYU, proposes an advanced hybrid solution, taking advantage of a gray-box model, utilizing a shallow neural network architecture to minimize the need for available labels of anomalies with the goal of designing a minimally-supervised and ultimately unsupervised anomaly detection algorithm, while maximizing robustness to noise and other artifacts.
Rick Parimi, https://www.linkedin.com/in/rickparimi/, Founder and CEO of Alexis Networks said, “We know that our One-Click Anomaly Detection™ ML software that is more efficient and complementary for tool business operators to find fraud and anomalies in any data set at the speed of business. But we are not resting on our laurels. We are aspiring to be the best AI and ML software company in the world – to be the best at Anomaly Detection. This R&D partnership with Assistant Professor S. Farokh Atashzar, https://engineering.nyu.edu/faculty/s-farokh-atashzar, at NYU Tandon shows our commitment to our vision and our passion to help customers overcome their largest and deepest data problems.”
Atashzar, an Assistant Professor of both Electrical and Computer, and Mechanical and Aerospace Engineering, will be leading this R&D project from NYU Tandon at MERIIT lab (https://wp.nyu.edu/meriit/).
“The fundamental algorithmic innovations and novelty in this project will have vast applicability for several applications in which availability of labels for anomalous events and the computational cost of conventional models are the concern. This project is supported by our extensive experience in the area of deep and shallow machine learning for data processing and anomaly detection at MERIIT lab – NYU,” said Atashzar, who is also affiliated with NYU Tandon’s Center for Urban Science and Progress and NYU WIRELESS.
