Abstract: A system and method for transferring an adversarial attack involving generating a surrogate model having an architecture and a dataset that mirrors at least one aspect of a target model of a target module, wherein the surrogate model includes a plurality of classes. The method involves generating a masked version of the surrogate model having fewer classes than the surrogate model by randomly selecting at least one class of the plurality of classes for removal. The method involves attacking the masked surrogate model to create a perturbed sample. The method involves generalizing the perturbed sample for use with the target module. The method involves transferring the perturbed sample to the target module to alter an operating parameter of the target model.

Abstract: An exemplary device for classifying an image includes a receiving unit that receives image data. The device also includes a hardware processor including a neural network architecture to extract a plurality of features from the image data, filter each feature extracted from the image data, concatenate the plurality of filtered features to form an image vector, evaluate the plurality of concatenated features in first and second layers of a plurality of fully connected layers of the neural network architecture based on an amount of deviation in the features determined at each fully connected layer, and generate a data signal based on an output of the plurality of fully connected layers. A transmitting unit sends the data signal to a peripheral or remote device.

Abstract: A computer-implemented method for generating an interpretable kernel embedding for heterogeneous data. The method can include identifying a set of base kernels in the heterogeneous data; and creating multiple sets of transformed kernels by applying a unique composition rule or a unique combination of multiple composition rules to the set of base kernels. The method can include fitting the multiple sets into a stochastic process model to generate fitting scores that respectively indicate a degree of the fitting for each of the multiple sets; storing the fitting scores in a matrix; and standardizing the matrix to generate the interpretable kernel embedding for the heterogeneous data.

Abstract: A system and method for transferring an adversarial attack involving generating a surrogate model having an architecture and a dataset that mirrors at least one aspect of a target model of a target module, wherein the surrogate model includes a plurality of classes. The method involves generating a masked version of the surrogate model having ewer classes than the surrogate model by randomly selecting at least one class of the plurality of classes for removal. The method involves attacking the masked surrogate model to create a perturbed sample. The method involves generalizing the perturbed sample for use with the target module. The method involves transferring the perturbed sample to the target module to alter an operating parameter of the target model.

Abstract: A method and system for cross-modal manifold alignment of different data domains includes determining for a shared embedding space a first embedding function for data of a first domain and a second embedding function for data of a second domain using a triplet loss, wherein triplets of the triplet loss include an anchor data point from the first, a positive and a negative data point from the second domain; creating a first mapping for the data of the first domain using the first embedding function in the shared embedding space; creating a second mapping for the data of the second domain using the second embedding function in the shared embedding space; and generating a cross-modal alignment for the data of the first domain and the data of the second domain.

Abstract: Exemplary systems and methods are directed to generating synthetic data for computer vision. A processing device generates a synthetic three-dimensional (3D) image of an object. A background image is selected, and a composite image is generated by combining the 3D image of the object and the background image. The processing device simulates: reflection or emission of at least one type of radiant energy from the surface of the object and/or the background according to a set of parameters associated with at least one of the object and the background image; and a reflectance or emittance measurement of the at least one type of radiant energy from the surface of the object by a sensor device configured for detecting the at least one type of radiant energy. The processing device generates a plurality of two-dimensional (2D) simulated images of different perspectives of the object based on simulation data.

Abstract: An exemplary device for classifying an image includes a receiving unit that receives image data. The device also includes a hardware processor including a neural network architecture to extract a plurality of features from the image data, filter each feature extracted from the image data, concatenate the plurality of filtered features to form an image vector, evaluate the plurality of concatenated features in first and second layers of a plurality of fully connected layers of the neural network architecture based on an amount of deviation in the features determined at each fully connected layer, and generate a data signal based on an output of the plurality of fully connected layers. A transmitting unit sends the data signal to a peripheral or remote device.

Abstract: An exemplary device for classifying an image includes a receiving unit that receives image data. The device also includes a hardware processor including a neural network architecture to extract a plurality of features from the image data, filter each feature extracted from the image data, concatenate the plurality of filtered features to form an image vector, evaluate the plurality of concatenated features in first and second layers of a plurality of fully connected layers of the neural network architecture based on an amount of deviation in the features determined at each fully connected layer, and generate a data signal based on an output of the plurality of fully connected layers. A transmitting unit sends the data signal to a peripheral or remote device.

Abstract: A method for detecting and/or identifying a cyber-attack on a network can include segmenting the network using a segmentation method with machine learning to generate one or more network segments; assigning a score to a data point within each network segment based on a presence or absence of an identified anomalous behavior of the data point; analyzing network data flow, via behavioral modeling, to provide a context for characterizing the anomalous behavior; combining, via a reinforcement learning agent, outputs of the segmentation method with behavioral modelling and assigned score to detect and/or identify a cyber-attack; providing one or more alerts to an analyst; receiving an analyst assessment of an effectiveness of the detection and/or identification; and providing the analyst assessment as feedback to the reinforcement learning agent.

Abstract: A system and method are disclosed for collecting and analyzing data in a cognitive fabric. The system can include a network of intelligent nodes, each node being configured for sharing or receiving data as a function of analytic processing to be performed at the node.

Abstract: A method for detecting and/or identifying a cyber-attack on a network can include segmenting the network using a segmentation method with machine learning to generate one or more network segments; assigning a score to a data point within each network segment based on a presence or absence of an identified anomalous behavior of the data point; analyzing network data flow, via behavioral modeling, to provide a context for characterizing the anomalous behavior; combining, via a reinforcement learning agent, outputs of the segmentation method with behavioral modelling and assigned score to detect and/or identify a cyber-attack; providing one or more alerts to an analyst; receiving an analyst assessment of an effectiveness of the detection and/or identification; and providing the analyst assessment as feedback to the reinforcement learning agent.

Abstract: A computer-implemented method for generating an interpretable kernel embedding for heterogeneous data. The method can include identifying a set of base kernels in the heterogeneous data; and creating multiple sets of transformed kernels by applying a unique composition rule or a unique combination of multiple composition rules to the set of base kernels. The method can include fitting the multiple sets into a stochastic process model to generate fitting scores that respectively indicate a degree of the fitting for each of the multiple sets; storing the fitting scores in a matrix; and standardizing the matrix to generate the interpretable kernel embedding for the heterogeneous data.

Abstract: A system and method are disclosed for collecting and analyzing data in a cognitive fabric. The system can include a network of intelligent nodes, each node being configured for sharing or receiving data as a function of analytic processing to be performed at the node.