Abstract: Techniques for auto tuning keepalive packets intervals to an optimal interval are described. A remote secure session between a client device and a server over a network is established. A determination is made to identify an optimal keepalive interval for sending packets to keep the remote secure session alive over the network, the optimal keepalive interval defining an amount of time between sending of packets that keep a connection open through middleboxes in the network. Keepalive test probes are transmitted by the client device and to the server at different time intervals. An optimal keepalive interval is determined based at least in part on the keepalive test probes transmitted at the different intervals. The client device transmits information indicating the optimal keepalive interval to the server. Finally, the client device transmits keepalive packets according to the optimal keepalive interval.

Abstract: Techniques and systems described herein relate to monitoring executions of computer instructions on computing devices based on learning and generating a control flow directed graph. The techniques and systems include determining a learned control flow directed graph for executable code of an application by observing executions of transitions during an observation period and determining destinations of indirect transfers based on the learned control flow directed graph. Next a disassembly of the executable code is determined based on the learned control flow directed graph, the destinations of the transfers, and the executable code.

Abstract: A system and method are provided for generating a cybersecurity behavioral graph from a log files and/or other telemetry data, which can be unstructured or semi-structured data. The log files are applied to a machine learning (ML) model (e.g., a large language model (LLM)) that generates/extract from the log files entities and relationships between said entities. The entities and relationships can be constrained using a cybersecurity ontology or schema to ensure that the results are meaningful to a cybersecurity context. A graph is then generated by mapping the extracted entities to nodes in the graph and the relationships to edges connecting nodes. To more efficiently extract the entities and relationships from the data file, an LLM is used to generate regular expressions for the format of the log files. Once generated, the regular expressions can rapidly parse the log files to extract the entities and relationships.

Abstract: A system and method are provided for generating a cybersecurity behavioral graph from a log files and/or other telemetry data, which can be unstructured or semi-structured data. The log files are applied to a machine learning (ML) model (e.g., a large language model (LLM)) that generates/extract from the log files entities and relationships between said entities. The entities and relationships can be constrained using a cybersecurity ontology or schema to ensure that the results are meaningful to a cybersecurity context. A graph is then generated by mapping the extracted entities to nodes in the graph and the relationships to edges connecting nodes. To more efficiently extract the entities and relationships from the data file, an LLM is used to generate regular expressions for the format of the log files. Once generated, the regular expressions can rapidly parse the log files to extract the entities and relationships.

Abstract: Techniques and systems described herein relate to monitoring executions of computer instructions on computing devices based on observing and generating a control flow directed graph. The techniques and systems include determining an observation phase for a process or application on a computing device. During the observation phase, CPU telemetry is determined and used to generate a control flow directed graph. After the control flow directed graph is generated, a hash table associated with frequently traversed execution paths is generated. A monitoring phase may be entered where transfers of instruction pointers are monitored based on the control flow directed graph to identify invalid transfers. The frequently traversed execution paths may be identified based on the hash table and be identified as valid if the hash value corresponds to the table.

Abstract: A system and method are provided for explaining ontological sub-graphs. The system and method include querying an ontology to determine a match between a query graph and a portion of an ontology graph. When there is a match, a subgraph representing the match is first translated into a simple summary using a simple language (e.g., triplets which include a subject and object corresponding to pairs of connected nodes in the subgraph and a verb/predicate representing a relation/edge in the subgraph that connect the pair nodes). This simple summary is then fed, as part of a prompt, to a large language model (LLM) that generates a human-readable summary based on the prompt.

Abstract: In one aspect, a method includes creating a polymorphic variant of a sample of malware, analyzing the polymorphic variant of the sample of malware by a security management service to determine if the polymorphic variant of the sample of malware evades detection by the security management service, when the security management service fails to detect the polymorphic variant during the analysis of the polymorphic variant, detonating the polymorphic variant in a virtualized environment to identify characterizations of the polymorphic variant, and training the security management service to detect the polymorphic variant based on the characterizations.

Abstract: A system and method are provided for generating a cybersecurity behavioral graph from a log files and/or other telemetry data, which can be unstructured or semi-structured data. The log files are applied to a machine learning (ML) model (e.g., a large language model (LLM)) that generates/extract from the log files entities and relationships between said entities. The entities and relationships can be constrained using a cybersecurity ontology or schema to ensure that the results are meaningful to a cybersecurity context. A graph is then generated by mapping the extracted entities to nodes in the graph and the relationships to edges connecting nodes. To more efficiently extract the entities and relationships from the data file, an LLM is used to generate regular expressions for the format of the log files. Once generated, the regular expressions can rapidly parse the log files to extract the entities and relationships.

Abstract: Techniques for auto tuning keepalive packets intervals to an optimal interval are described. A remote secure session between a client device and a server over a network is established. A determination is made to identify an optimal keepalive interval for sending packets to keep the remote secure session alive over the network, the optimal keepalive interval defining an amount of time between sending of packets that keep a connection open through middleboxes in the network. Keepalive test probes are transmitted by the client device and to the server at different time intervals. An optimal keepalive interval is determined based at least in part on the keepalive test probes transmitted at the different intervals. The client device transmits information indicating the optimal keepalive interval to the server. Finally, the client device transmits keepalive packets according to the optimal keepalive interval.

Abstract: A method of defining priority of a number of data packets within a queue includes generating a policy. The policy defines a first multiplexed channel of a plurality of multiplexed channels. The first multiplexed channel having a first priority. The policy also defines a second multiplexed channel of the plurality of multiplexed channels. The second multiplexed channel having a second priority. The first priority is defined as being of a higher priority relative to the second priority. The method further includes receiving the number of data packets over the plurality of multiplexed channels associated with a session based at least in part on the policy.

Abstract: Techniques and systems described herein relate to monitoring executions of computer instructions on computing devices based on learning and generating a control flow directed graph. The techniques and systems include determining an observation phase for a process or application on a computing device. During the observation phase, CPU telemetry is determined and used to generate a control flow directed graph. After the control flow directed graph is generated, a monitoring phase may be entered where transfers of instruction pointers are monitored based on the control flow directed graph to identify invalid transfers.

Abstract: Techniques and systems described herein relate to monitoring executions of computer instructions on computing devices based on learning and generating a control flow directed graph. The techniques and systems include determining a learned control flow directed graph for executable code of an application by observing executions of transitions during an observation period and determining destinations of indirect transfers based on the learned control flow directed graph. Next a disassembly of the executable code is determined based on the learned control flow directed graph, the destinations of the transfers, and the executable code.

Abstract: Techniques and systems described herein relate to monitoring executions of computer instructions on computing devices based on observing and generating a control flow directed graph. The techniques and systems include determining an observation phase for a process or application on a computing device. During the observation phase, CPU telemetry is determined and used to generate a control flow directed graph. After the control flow directed graph is generated, a monitoring phase may be entered where transfers of instruction pointers are monitored based on the control flow directed graph to identify invalid transfers. Transition to the monitoring phase may be based on determining a confidence score in the observed control flow directed graph and causing the transition when the confidence score is above a threshold.

Abstract: Techniques and systems described herein relate to monitoring executions of computer instructions on computing devices based on learning and generating a control flow directed graph. The techniques and systems include determining a learned control flow directed graph for a process executed on the computing system. A system call is identified during execution of the process as well as a predetermined number of transitions leading to the system call. A validity of the transitions leading the system call is determined based on the learned control flow directed graph and the computing system may perform an action based on the validity.

Abstract: Techniques and systems described herein relate to monitoring executions of computer instructions on computing devices based on learning and generating a control flow directed graph. The techniques and systems include determining a learned control flow diagram for a process on a computing system and monitoring execution of the process on the computing system using the control flow diagram. An unobserved transition is determined based on the learned control flow diagram and the unobserved transition is classified as safe or unsafe based on a monitoring component analysis. An action is performed based on the safety classification and the learned control flow diagram.

Abstract: Techniques and systems described herein relate to monitoring executions of computer instructions on computing devices based on learning and generating a control flow directed graph. The techniques and systems include determining telemetry representing execution of a process on a computing system and accessing a learned control flow diagram graph for the process. A transfer of an instruction pointer is determined based on the telemetry and a validity of the transfer is determined based on the learned control flow directed graph. If invalid, then an action to terminate the process is determined, otherwise the action may be allowed to execute when valid.

Abstract: A method of defining priority of a number of data packets within a queue includes generating a policy. The policy defines a first multiplexed channel of a plurality of multiplexed channels. The first multiplexed channel having a first priority. The policy also defines a second multiplexed channel of the plurality of multiplexed channels. The second multiplexed channel having a second priority. The first priority is defined as being of a higher priority relative to the second priority. The method further includes receiving the number of data packets over the plurality of multiplexed channels associated with a session based at least in part on the policy.

Abstract: A trained model may be deployed to an Internet-of-Things (IOT) operational environment in order to ingest features and detect events extracted from network traffic. The model may be received and converted into a meta-language representation which is interpretable by a data plane engine. The converted model can then be deployed to the data plane and may extract features from network communications over the data plane. The extracted features may be fed to the deployed model in order to generate event classifications or device state classifications.

Abstract: A trained model may be deployed to an Internet-of-Things (IOT) operational environment in order to ingest features and detect events extracted from network traffic. The model may be received and converted into a meta-language representation which is interpretable by a data plane engine. The converted model can then be deployed to the data plane and may extract features from network communications over the data plane. The extracted features may be fed to the deployed model in order to generate event classifications or device state classifications.