Abstract: Computer vision systems and methods for object detection with reinforcement learning are provided. The system includes a reinforcement learning agent configured to detect an object pertaining to a target object class and a plurality of objects pertaining to different target object classes, such that the reinforcement learning agent determines a bounding box for each of the detected of objects. The system first sets parameters of the reinforcement learning agent. The system then detects an object and/or objects in an image based on the set parameters. Finally, the system determines a bounding box and/or bounding boxes for each of the detected objects.

Abstract: An apparatus is provided, including a transaction card, having: a first microprocessor configured to exchange transaction data; a sensor configured to detect movement; and a second microprocessor in communication with the sensor. The second microprocessor may be configured to store arrays of hash values, representing a movement of a user and a card; obtain hash values based on observed movement patterns, and compare them to the stored values. Upon receiving a transaction request, the microprocessor may determine in real time whether the user is in locomotion and whether the card is being intentionally tapped by the user, based on the observed movement patterns. On this basis, the second microprocessor may determine whether to require an augmented authentication for the requested transaction.

Abstract: A method includes receiving a request. The request is encrypted to generate an encrypted request. The encrypted request includes a version information for a first software package deployed in a standalone computing system, and a region information and a hardware information for the standalone computing system. The encrypted request is sent to a respective regional data center of a cloud computing system based on the region information of the first software package. A second software package is identified among software packages stored in the respective regional data center. The second software package corresponds to the hardware information of the first software package. The version information of the first software package is compared to a version information of the second software package. In response to identifying that the second software package is more recent than the first software package, the second software package is deployed to the standalone computing system.

Abstract: Bias supplies and plasma processing systems are disclosed. One bias supply comprises an output node, a return node, and a switch network and at least one power supply coupled to the output node and the return node. The switch network and the at least one power supply configured, in combination, to apply an asymmetric periodic voltage waveform and provide a corresponding current waveform at the output node relative to the return node. A timing parameter estimator receives a digital representation of a full cycle of the voltage and current waveforms, and generates a pulse width control signal based on a crossing time that the current waveform crosses a threshold current value after falling from a positive peak current value to control the switch network.

Abstract: An apparatus comprises a memory communicatively coupled to a processor. The memory is configured to store network interchange parameters and user profiles. The processor is configured to receive a data exchange request triggered by a user device. Further, the processor is configured to request, to a decentralized network, a token indicating that a user profile associated with the user device is entitled to access an entitlement, determine whether a data exchange network is available to exchange data in accordance with the network interchange parameters, identify a second data exchange network that is available in accordance with second network interchange parameters in response to determine that the data exchange network is unavailable, and request a data exchange output from the second data exchange network via the token. The processor is configured to receive the data exchange output from the second data exchange network.

Abstract: A color-changing smart card may be provided. The smart card may include a first layer encasing at least a front of the smart card. The smart card may include a smart chip on the front of the smart card. The smart chip may be configured for transmitting and receiving electronic communications to and from a point of sale (“POS”) device. The smart card may include microcapsules residing on a second layer of the smart card. When the microcapsules are in a rest state the first layer may have a first color. When the microcapsules are in an active state the first layer may have a second color. The microcapsules may be triggered to change the color in response to the smart chip receiving an electronic communication of a status of a transaction initiated by the smart card at the POS device.

Abstract: Aspects of the disclosure relate to reimaging an ATM using an extended reality interface and machine learning models that are configured to reimage the ATM. A computing system may send a request for authorization to access an ATM. Based on the request being granted the ATM may be accessed via an extended reality interface. Reimaging the ATM may be based on inputting ATM data associated with a state of the ATM into machine learning models that are configured to generate instructions to reimage the ATM and then execute threads based on processes that were determined based on the instructions. The threads may be executed in a RAM of the ATM in accordance with machine learning model determined priorities of the threads.

Abstract: A system for reestablishing network connectivity for particular network equipment in a failed network zone comprises a processor associated with a server. The processor receives a flag that identifies the failed network zone including particular network equipment that have lost network connectivity in a network. The processor generates a first non-fungible token using the configuration information associated with each of the particular network equipment. The processor determines routing information based at least in part upon the configuration information and a network topology. The processor generates a second non-fungible token based at least in part upon the routing information. The processor aggregates the first non-fungible token and the second non-fungible token to generate a primary non-fungible token. The processor reestablishes network connectivity for at least a portion of the particular network equipment based at least in part upon information associated with the primary non-fungible token.

Abstract: A computing chip may comprise a memory that stores a machine-learning model. an optical absorption spectroscopy component, and one or more processors operably coupled to the memory and the optical absorption spectroscopy component. The one or more processors may cause the optical absorption spectroscopy component to emit radiation to N pre-determined directions within an internal space of a computing device. The one or more processors may cause the optical absorption spectroscopy component to sample an absorption of reflections of the emitted radiation for each of the N pre-determined directions. The one or more processors may determine that one or more hardware components of the computing device are in particular abnormal conditions by processing the N sampled absorptions with the machine-learning model. The one or more processors may send a notification indicating that the one or more hardware components are in the particular abnormal conditions.

Abstract: Aspects of this disclosure relate to systems and methods to determine whether an application is misconfigured or malicious using smart contracts stored on a distributed ledger and a transferred deep learning system. A smart contract is generated and assigned rules for monitoring an application session on a user computing device when an application session associated with the application is initiated. The smart contract denies or grants the application session permission to access user data and shuts down the application session when a malicious indication is detected in configuration data associated with the application based on rules assigned to the smart contract. The malicious indication is output by a deep learning module trained on historical configuration data associated with the application. The deep learning module transfers layers to an enterprise deep learning module on an enterprise platform to determine a misconfiguration indication indicating the configuration data is misconfigured.

Abstract: Arrangements for remote device control and transaction processing are provided. In some aspects, a local computing device may initiate a communication session with the self-service kiosk. The local computing device may capture self-service kiosk identifying data and may generate a deep link to an application. The local computing device may transmit the deep link and self-service kiosk identifying information to a remote computing device. An indication of selection of the deep link may be received by the local computing device and a remote transaction processing session may be initiated at the self-service kiosk. The self-service kiosk may transmit a user interface for display by the remote computing device. An option from the user interface may be selected for processing. The self-service kiosk may process the selected option, send a notification of transaction completion to the remote user computing device and terminate the communication session with the local user computing device.

Abstract: A system for reestablishing network connectivity for particular network equipment in a failed network zone comprises a processor associated with a server. The processor receives a flag that identifies the failed network zone including particular network equipment that have lost network connectivity in a network. The processor generates a first non-fungible token using the configuration information associated with each of the particular network equipment. The processor determines routing information based at least in part upon the configuration information and a network topology. The processor generates a second non-fungible token based at least in part upon the routing information. The processor aggregates the first non-fungible token and the second non-fungible token to generate a primary non-fungible token. The processor reestablishes network connectivity for at least a portion of the particular network equipment based at least in part upon information associated with the primary non-fungible token.

Abstract: A smart-dust information security and energy optimization process/machine/system is disclosed for use with an ATM. Tiny microelectromechanical systems (MEMS) less than 1 mm in size are deployed and distributed in and around the ATM internal and external components. The MEMS include various sensors to detect component malfunctions and identify security threats (e.g., card skimmers), and also to place the ATM in a reduced energy state whenever possible. MEMS interoperable components and sensors may include, inter alia, batteries, wireless transmitters/receivers, processors, memory, analog-to-digital converters, LIDAR, photo detectors, vibration detectors, motion detectors, wireless signal detectors for detecting Bluetooth or RFID signals, temperature sensors, humidity sensor, a power consumption/induced electromagnetic energy sensors, piezoelectric sensors, sound sensors, etc.

Abstract: Apparatus and methods for a li-fi payment network are provided. A customer's device may communicate, directly or through intermediate devices and computer systems, over a li-fi network with a point-of-sale terminal. A bill may be transmitted to the customer's device over the li-fi network. The customer may select an option to pay the bill. The selection, along with information to authorize and process the payment, may be transmitted over the li-fi network to the POS terminal. The POS terminal may process the payment and generate a receipt. The receipt may be transmitted to the customer's device over the li-fi network.

Abstract: An electronic bill may be generated, transmitted, and paid remotely at a POS using a UWB-compatible user device. The location of the user device may be detected leveraging UWB technology at the POS. The bill may be pushed to the user device over a MEMS network of sensor nodes at the POS. An optimal route over the sensor nodes may be selected so that the electronic bill is directed to a sensor node in proximity to the user device. The selected route may minimize the number of sensor nodes over which the bill must pass. The route may be dynamically changed when the user device is moved from one location to another at the POS. The customer may specify an electronic payment method in response to receiving the bill or the customer may have pre-specified the payment method. An electronic confirmation of payment may be transmitted to the POS.

Abstract: Computer vision systems and methods for end-to end training of neural networks are provided. The system generates a fixed point algorithm for dual-decomposition of a maximum-a-posteriori inference problem and trains the convolutional neural network and a conditional random field with the fixed point algorithm and a plurality of images of a dataset to learn to perform semantic image segmentation. The system can segment an attribute of an image of the dataset by the trained neural network and the conditional random field.

Abstract: A decentralized swarm intelligence algorithm over a network of paired ATMs to prevent or control surface attacks on ATMs. The ATMs may be seeded with an initial swarm intelligence model to identify suspicious activity. An ATM may relay alerts about suspicious activity at that ATM to other paired ATMs. The ATMs use machine learning to update the model and perform swarm intelligence autonomously at the paired ATMs. A bank may provide the initial swarm intelligence model, receive alerts about and analyze the attacks, and provide updated swarm intelligence models to prevent or limit future attacks. Smart contracts may be used to specify rules for performing swarm intelligence using the paired ATMs. Records regarding swarm intelligence models, attempted suspicious attacks including payment instruments that may have been used, actions taken in response to the attacks, and smart contracts may be recorded on a blockchain distributed ledger.

Abstract: An ATM includes a long range WAN module that provides the ATM with a low power WAN (LPWAN) capability for increased resiliency. The ATM and bank application servers generally communicate over an ATM communications network to allow customers to perform banking transactions. When this connection is not functioning, the ATM may continue to be used to perform banking transactions, without interruption, by communicating over a LoRaWAN network with another nearby LPWAN-enabled ATM that is in communication with the bank application servers. The LoRaWAN network may also be used to automatically obtain software to fix an error at the ATM when the software is obtainable via a nearby LPWAN-enabled ATM. The LoRaWAN module may be further used to record a state of operations at the ATM and to use the record to resume the transaction where the transaction was interrupted while in progress due to an ATM outage.

Abstract: Computer vision systems and methods for object detection with reinforcement learning are provided. The system includes a reinforcement learning agent configured to detect an object pertaining to a target object class and a plurality of objects pertaining to different target object classes, such that the reinforcement learning agent determines a bounding box for each of the detected of objects. The system first sets parameters of the reinforcement learning agent. The system then detects an object and/or objects in an image based on the set parameters. Finally, the system determines a bounding box and/or bounding boxes for each of the detected objects.

Abstract: Computer vision systems and methods for text classification are provided. The system detects a plurality of text regions in an image and generates a bounding box for each detected text region. The system utilizes a neural network to recognize text present within each bounding box and classifies the recognized text, based on at least one extracted feature of each bounding box and the recognized text present within each bounding box, according to a plurality of predefined tags. The system can associate a key with a value and return a key-value pair for each predefined tag.