Abstract: A method of authenticating a received signal comprises receiving signals within a spectrum of bandwidth B, converting the signals into a frequency domain representation of bandwidth B, and segmenting the frequency domain representation of bandwidth B into spectrum segments each of training bandwidth W. For each spectrum segment, the method further comprises applying the spectrum segment to a neural network, and receiving, from the neural network, a radio frequency identification map comprising two or more classification vectors. The method further comprises determining a total variation across the two or more classification vectors and identifying the received signal as authenticated when the total variation is less than a threshold value. The method may further comprise training the neural network using a training dataset of wideband training samples. The method may further comprise generating each of the wideband training samples by stitching together two or more stored signals.

Abstract: A method of protecting a neural network from an adversarial attack, comprises generating a first context, generating, by the hypernetwork, a first set of weights based on the first context, and applying the first set of weights to the neural network. The method may further comprise generating a second context subsequent to the first context, generating, by the hypernetwork, a second set of weights based on the second context, and applying the second set of weights to the neural network. The method may further comprise generating additional contexts and corresponding sets of weights, and subsequently applying each of the corresponding sets of weights to the neural network.

Abstract: Systems and method of classification are provided. Upon receiving an input, a feature set is defined from the input. A semantic cluster to be associated with the input is defined based on the feature set, the semantic cluster being one of a plurality of semantic clusters each defining a subset of outputs of a neural network based on semantic similarity of the subset. The feature set is applied to a subgraph corresponding to the semantic cluster, the subgraph being one of a plurality of subgraphs each defining a portion of the neural network. A classification for the input is then be determined based on an output of the subgraph.

Abstract: Adapters for Deep Neural Networks (DNNs) are described for use in the presence of corruption/noise affecting inputs. These adapters can proactively learn latent representations of different corruption types, each associated with a sub-network state tailored to correctly classify inputs affected by that corruption. After deployment, they adapt the DNN to previously unseen corruptions in an unsupervised fashion by (i) estimating the latent representation of the ongoing corruption; (ii) selecting the sub-network whose associated corruption is the closest in the latent space to the ongoing corruption; and (iii) adapting the DNN state, so that its representation matches the ongoing corruption. Resource-efficient and swift adaptation to new distributions caused by different corruptions can be achieved without requiring a large variety of input data.

Abstract: A method of generating a deep neural network (DNN) may comprise receiving one or more application-level requirements associated with network communications, translating the one or more application-level requirements into one or more technical constraints, and providing the one or more technical constraints to a control loop that generates a certified DNN architecture based on the technical constraints. The control loop may further comprise a DNN search engine and a hardware synthesis engine. The method may comprise selecting, using the DNN search engine, a candidate DNN architecture based on the technical constraints, and generating, using the hardware synthesis engine, a hardware architecture corresponding to the selected candidate DNN architecture.

Abstract: Systems and associated methods for detecting a set of spectrum sensing falsification (SSF) attacks in a geographic database (GDB) driven cognitive radio (CR) system. Viewing the GDB as a type of non-orthogonal compressive sensing (CS) dictionary, the composite power spectral density (PSD) estimate at a candidate CR is approximated by a small number of sensor nodes listed in the GDB. In a dense CR network, the PSD estimate at a CR may contain a composite mixture of spectrally overlapping signals. An implementation of an optimized, greedy algorithm orthogonal matching pursuit (OMP) returns a set of sensor nodes which are suspected to be in the vicinity of the CR. A sufficient match between the PSD estimate reported by a candidate CR and the PSD that is sparsely approximated from the SNs in its area provides confidence (trust) metrics which may be used to detect potential SSF attacks.

Abstract: The use of shielded material in a deployable vehicle arresting and containment device that, when used for the interception of an unmanned vehicle, effectively achieves RF isolation of that vehicle, breaking all external communications with that vehicle. This apparatus, which may have internal and external antennas, could enable a variety of advanced effects such as localized GPS and command and control link spoofing and jamming as well as providing a vehicle for signal intercept and intelligence solutions. Additionally, due to the shielding properties of the arresting and containment device, semi-destructive means such as localized EMPs could be used to damage the encapsulated unmanned vehicle electronics.

Abstract: The use of shielded material in a deployable vehicle arresting and containment device that, when used for the interception of an unmanned vehicle, effectively achieves RF isolation of that vehicle, breaking all external communications with that vehicle. This apparatus, which may have internal and external antennas, could enable a variety of advanced effects such as localized GPS and command and control link spoofing and jamming as well as providing a vehicle for signal intercept and intelligence solutions. Additionally, due to the shielding properties of the arresting and containment device, semi-destructive means such as localized EMPs could be used to damage the encapsulated unmanned vehicle electronics.