Citibank, N.A.

The systems and methods disclosed herein monitor application (e.g., artificial intelligence (AI) model) operations using interactions between the application and a kernel. The systems and methods disclosed herein intercept, using a kernel interface, one or more function invocations transmitted from the application (e.g., an AI model without model modification). Event record(s) are generated for one or more functions to define process identifiers, resource interaction types, timestamps, and/or resource identifiers. Observed pattern(s) for the application are identified by comparing current event record(s) with previous record(s), and the identified observed pattern(s) are evaluated against reference pattern(s) to generate score(s). Data packet(s) that indicate observed pattern(s), corresponding score(s), and/or cryptographic digital fingerprint(s) of the one or more functions are generated. The data packet(s) are transmitted to distributed ledgers for immutable storage.

The systems and methods disclosed herein monitor application (e.g., artificial intelligence (AI) model) operations using interactions between the application and a kernel. The systems and methods disclosed herein intercept, using a kernel interface, one or more function invocations transmitted from the application (e.g., an AI model without model modification). Event record(s) are generated for one or more functions to define process identifiers, resource interaction types, timestamps, and/or resource identifiers. Observed pattern(s) for the application are identified by comparing current event record(s) with previous record(s), and the identified observed pattern(s) are evaluated against reference pattern(s) to generate score(s). Data packet(s) that indicate observed pattern(s), corresponding score(s), and/or cryptographic digital fingerprint(s) of the one or more functions are generated. The data packet(s) are transmitted to distributed ledgers for immutable storage.

The present disclosure provides a malfunction interdiction system with monitoring, detection, and response capabilities specifically designed for distributed AI environments. The system may first detect any AI agents that are malfunctioning and then remediate any malfunctions via quarantine and/or AI agent termination. Upon detection of a malfunction, the system may automatically isolate the compromised agent through a quarantine process that modifies network configuration settings to block communication with network devices while maintaining connectivity to monitoring systems for forensic analysis and potential remediation. The system may additionally or alternatively initiate graduated deactivation procedures ranging from partial restrictions to full deactivation.

Presented herein are system and methods for detecting anomalies in microservices. A server may receive a first plurality of metrics from a microservice of a plurality of microservices. Each of the microservices may provide a respective function for an application independently from other microservices. The server may apply the first plurality of metrics to an ensemble of anomaly detection models to generate classifications. Each of the classifications may indicate the first plurality of metrics as one of anomalous or normal from a respective model of the ensemble. The ensemble may be trained using a second plurality of metrics from the microservice. The server may identify a majority of the plurality of classifications as indicating the first plurality of metrics as anomalous. The server may determine that at least one of the first plurality of metrics satisfies a threshold. The server may generate an alert to indicate an anomaly event.

The systems and methods disclosed herein monitor application (e.g., artificial intelligence (AI) model) operations using interactions between the application and a kernel. The systems and methods disclosed herein intercept, using a kernel interface, one or more function invocations transmitted from the application (e.g., an AI model without model modification). Event record(s) are generated for one or more functions to define process identifiers, resource interaction types, timestamps, and/or resource identifiers. Observed pattern(s) for the application are identified by comparing current event record(s) with previous record(s), and the identified observed pattern(s) are evaluated against reference pattern(s) to generate score(s). Data packet(s) that indicate observed pattern(s), corresponding score(s), and/or cryptographic digital fingerprint(s) of the one or more functions are generated. The data packet(s) are transmitted to distributed ledgers for immutable storage.

Systems and methods for analyzing operations to determine factors contributing to a malicious classification and providing recommendations to prevent future misclassifications are disclosed herein. A feature modification system receives operation data associated with monitored operations, where each operation is characterized by a set of feature values. A machine learning-based detection model processes the operation data to generate a prediction indicating whether each operation is malicious. Based on a prediction indicating that an operation has been classified as malicious, the system determines the impact of each feature on the prediction as well as the variability of each feature to identify impactful and modifiable features. The system generates an input dataset including entries with modified features and processes the dataset to obtain a new set of predictions. If a modified entry results in a non-malicious classification, the system generates a recommendation indicating the modifications required to avoid future misclassifications.

Presented herein are system and methods for countermeasures to address anomalies in microservices. A server having one or more processors coupled with memory may receive a first plurality of metrics from a defined set of microservices for a function. The server may apply the first plurality of metrics to an ensemble of anomaly detection models to generate a plurality of classifications. Each classification may indicate the first plurality of metrics as one of anomalous or normal from a respective model of the ensemble of anomaly detection models. The server may identify a majority of the plurality of classifications as corresponding to an anomaly event. The server may determine, responsive to identifying the majority, that at least one of the first plurality of metrics satisfies a criterion of a policy of a plurality of policies. The server may perform a countermeasure identified by the policy to address the anomaly event.

Systems and methods for reducing network traffic by filtering network requests based on network request-related information to be transmitted to one or more remote computing systems are disclosed. The system receives a first network operation indicating (i) a request to access a first resource and (ii) a set of requestor specific criteria associated with accessing the first resource. The system identifies a set of entities associated with the first resource and selectively communicates with a filtered subset of the set of entities by: identifying a set of entity specific criteria for each entity of the set of entities; determining whether the requestor specific criteria satisfies the set of entity specific criteria of respective entities; and transmitting the first network operation to respective entities in response to the requestor specific criteria satisfying the set of entity specific criteria of the respective entity.

Presented herein are system and methods for countermeasures to address anomalies in microservices. A server having one or more processors coupled with memory may receive a first plurality of metrics from a defined set of microservices for a function. The server may apply the first plurality of metrics to an ensemble of anomaly detection models to generate a plurality of classifications. Each classification may indicate the first plurality of metrics as one of anomalous or normal from a respective model of the ensemble of anomaly detection models. The server may identify a majority of the plurality of classifications as corresponding to an anomaly event. The server may determine, responsive to identifying the majority, that at least one of the first plurality of metrics satisfies a criterion of a policy of a plurality of policies. The server may perform a countermeasure identified by the policy to address the anomaly event.

Systems and methods for generating parameters for detecting malicious activity using decision trees are disclosed herein. The system generates parameters using a decision tree having predictor features and a target feature. The system generates, in near real time as messages are received, outputs for the target feature based on the parameters. The system determines, after a period of time, that a primary value of a primary external metric exceeds a primary threshold. The primary value may be indicative of the outputs having insufficient accuracy. The system determines a new decision tree that includes predictor features relating to the primary external metric. The new decision tree may also generate outputs that indicate that the primary value of the primary external metric is less than the primary threshold. The system generates new parameters using the new decision tree and uses the new parameters to generate new outputs for the target feature.