Abstract: A cognitive security analytics platform is enhanced by providing a technique for automatically inferring temporal relationship data for cybersecurity events. In operation, a description of a security event is received, typically as unstructured security content or data. Information such as temporal data or cues, are extracted from the description, along with security entity and relationship data. Extracted temporal information is processing according to a set of temporal markers (heuristics) to determine a time value marker (i.e., an established time) of the security event. This processing typically involves retrieval of information from one or more structured data sources. The established time is linked to the security entities and relationships. The resulting security event, as augmented with the identified temporal data, is then subjected to a management operation.

Abstract: An interactive display system enables a user to compose a graph pattern for a temporal graph on a display screen. The system comprises a canvas that provides an interactive editing surface. The editor receives an input a set of user interactions, such as the drawing of lines and boxes, the specifying of attributes, and the like, that together compose a graph pattern. During the graph pattern composition, the user may retrieve other graph patterns (e.g., from a data store) and integrate them into the pattern being composed. Once the graph pattern is composed (or as it is being composed), the system converts the graphical pattern into a text-based representation, such as a computer program in a particular graph programming language, which is then used for subsequent processing and matching in a cybersecurity threat discovery workflow. The pattern (program code) also is stored to disk, from which it may be retrieved and converted back into its graphical view on the screen, e.g., for further editing and revision.

Abstract: A method to detect anomalous behavior in an execution environment. A set of system events captured from a monitored computing system are received. Using the received system events, a model is then trained using machine learning. The model is trained to automatically extract one or more features for the received set of system events, wherein a system event feature is determined by a semantic analysis and represents a semantic relationship between or among a grouping of system events that are observed to co-occur in an observation sample. An observation sample is associated with an operating scenario that has occurred in the execution environment. Once trained, and using the features, the model is used to detect anomalous behavior. As an optimization, prior to training, the set of system events are pre-processed into a reduced set of system events. The modeler may comprise a component of a malware detection system.

Abstract: An automated technique for security monitoring leverages a labeled semi-directed temporal graph derived from system-generated events. The temporal graph is mined to derive process-centric subgraphs, with each subgraph consisting of events related to a process. The subgraphs are then processed to identify atomic operations shared by the processes, wherein an atomic operation comprises a sequence of system-generated events that provide an objective context of interest. The temporal graph is then reconstructed by substituting the identified atomic operations derived from the subgraphs for the edges in the original temporal graph, thereby generating a reconstructed temporal graph. Using graph embedding, the reconstructed graph is converted into a representation suitable for further machine learning, e.g., using a deep neural network. The network is then trained to learn the intention underlying the temporal graph.

Abstract: A computer system trains an Artificial Intelligence (AI) model to generate a key generated as a same key based on multiple different feature vectors, which are based on specified target environment attributes of a target environment domain. The computer system uses the key to encrypt concealed information as an encrypted payload and distributes the encrypted payload and the trained AI model to another computer system. The other computer system extracts environment attributes based on an environment domain accessible by the other computer system and decodes a candidate key by using the trained AI model that uses the extracted environment attributes of the environment domain as input. The trained AI model is trained to generate a key that is generated as a same key from multiple different feature vectors corresponding to specified target environment attributes of a target environment domain. The other computer system determines whether the candidate key is correct.

Abstract: Unknown and reference signatures are accessed. The unknown and reference signatures indicate patterns that correspond to known threats to resources (such as computer systems and/or computer networks) in a computer environment and comprise a multitude of descriptive elements having information describing different aspects of a corresponding signature. A set of similarity measures is created of the unknown and reference signatures from different perspectives, each perspective corresponding to a descriptive element. The set of similarity measures are integrated to generate an overall similarity metric. The overall similarity metric is used to find appropriate categories in the reference signatures into which the unknown signatures should be placed. The unknown signatures are placed into the appropriate categories to create a mapping from the unknown signatures to the reference signatures.

Abstract: Unknown and reference signatures are accessed. The unknown and reference signatures indicate patterns that correspond to known threats to resources (such as computer systems and/or computer networks) in a computer environment and comprise a multitude of descriptive elements having information describing different aspects of a corresponding signature. A set of similarity measures is created of the unknown and reference signatures from different perspectives, each perspective corresponding to a descriptive element. The set of similarity measures are integrated to generate an overall similarity metric. The overall similarity metric is used to find appropriate categories in the reference signatures into which the unknown signatures should be placed. The unknown signatures are placed into the appropriate categories to create a mapping from the unknown signatures to the reference signatures.

Abstract: A decoy filesystem that curtails data theft and ensures file integrity protection through deception is described. To protect a base filesystem, the approach herein involves transparently creating multiple levels of stacking to enable various protection features, namely, monitoring file accesses, hiding and redacting sensitive files with baits, and injecting decoys onto fake system views that are purveyed to untrusted subjects, all while maintaining a pristine state to legitimate processes. In one implementation, a kernel hot-patch is used to seamlessly integrate the new filesystem module into live and existing environments.

Abstract: A cloud based implemented method (and apparatus) includes receiving input data including bipartite graph data in a format of source MAC (Media Access Control) address data versus destination IP (Internet Protocol) data and timestamp information, and providing the input bipartite graph data into a first processing to detect malicious beaconing activities using a lockstep detection module on the input bipartite graph data, as executed in a cloud environment, to detect possible synchronized attacks against a targeted infrastructure.

Abstract: Decoy network ports and services are projected onto existing production workloads to facilitate cyber deception, without the need to modify production machines. The approach may be implemented in a production network that includes two segments. A production machine is reachable via the first segment, while a decoy machine that offers the network service expected from the production machine is reachable via the second segment. A deception router is configured in front of the two segments, and it is not visible on the link and network layers. The router inspects network traffic destined for the production machine. Based on a set of one or more conditions being met, the router determines whether to relay network packets to the production machine, or to redirect the packet to the decoy machine.

Abstract: This disclosure provides an automatic causality tracking system that meets real-time analysis needs. It solves causality tracking for cybersecurity, preferably as three sub-tasks: backward tracking, forward tracking, and path-finding. Given a set of threat indicators, the first sub-task yields the system elements (e.g., entities such as processes, files, network sockets, and the like) that contribute information to a set of threat indicators backward in time. The second sub-task yields system elements forward in time. Given two sets of threat indicators, the third sub-task yields shortest paths between them, e.g., how the two sets of indicators are connected to one another. The system enables efficient multi-point traversal analysis with respect to a set of potential compromise points, and using data from real information flows.

Abstract: An automated method for processing security events in association with a cybersecurity knowledge graph. The method begins upon receipt of information from a security system representing an offense. An initial offense context graph is built based in part on context data about the offense. The graph also activity nodes connected to a root node; at least one activity node includes an observable. The root node and its one or more activity nodes represent a context for the offense. The knowledge graph, and potentially other data sources, are then explored to further refine the initial graph to generate a refined graph that is then provided to an analyst for further review and analysis. Knowledge graph exploration involves locating the observables and their connections in the knowledge graph, determining that they are associated with known malicious entities, and then building subgraphs that are then merged into the initial graph.

Abstract: Rapid deployments of application-level deceptions (i.e., booby traps) implant cyber deceptions into running legacy applications both on production and decoy systems. Once a booby trap is tripped, the affected code is moved into a decoy sandbox for further monitoring and forensics. To this end, this disclosure provides for unprivileged, lightweight application sandboxing to facilitate monitoring and analysis of attacks as they occur, all without the overhead of current state-of-the-art approaches. Preferably, the approach transparently moves the suspicious process to an embedded decoy sandbox, with no disruption of the application workflow (i.e., no process restart or reload). Further, the action of switching execution from the original operating environment to the sandbox preferably is triggered from within the running process.

Abstract: A computer-implemented method (and apparatus) includes receiving input data comprising bipartite graph data in a format of source MAC (Machine Access Code) data versus destination IP (Internet Protocol) data and timestamp information. The input bipartite graph data is provided into a first processing to detect malicious beaconing activities using a lockstep detection method on the input bipartite graph data to detect possible synchronized attacks against a targeted infrastructure. The input bipartite graph data is also provided into a second processing, the second processing initially converting the bipartite graph data into a co-occurrence graph format that indicates in a graph format how devices in the targeted infrastructure communicate with different external destination servers over time. The second processing detects malicious beaconing activities by analyzing data exchanges with the external destination servers to detect anomalies.

Abstract: Advanced Persistent Threat (APT) defense leverages priority-based tracking around alerts, together with priority-based alert reasoning task scheduling. In one embodiment, individual alert reasoning tasks are managed by an alert scheduler, which effectively allocates available computation resources to prioritize the alert reasoning tasks, which each execute within processing workers that are controlled by the alert scheduler. An alert reasoning task typically is prioritized (relative to other such tasks) according to one or more factors, such as severity levels, elapsed time, and other tracking results. By implementing priority-based task scheduling, the task scheduler provides for alert reasoning tasks that are interruptible. In this approach, and once an alert is assigned to a task and the task assigned a worker, priority-based connectivity tracker around each alert is carried out to provide further computational efficiency.

Abstract: A command endpoint used by Domain Generation Algorithm (DGA) malware is identified using machine learning-based clustering. According to this technique, at least one attribute associated with a candidate resolved DNS name is identified. The candidate resolved DNS name has associated therewith a set of names that are failed DNS lookups but that cluster with the candidate resolved DNS name. A set of additional names that share the at least one attribute with the candidate resolved DNS name are then identified. For the set of additional names, an extent to which the set of additional names also clusters with the set of names that are failed DNS lookups is then determined. The candidate resolved DNS name is characterized as associated with the command endpoint when the set of additional names cluster with the set of names that are failed DNS lookups to a configurable degree.

Abstract: A cognitive security analytics platform is enhanced by providing a technique for automatically inferring temporal relationship data for cybersecurity events. In operation, a description of a security event is received, typically as unstructured security content or data. Information such as temporal data or cues, are extracted from the description, along with security entity and relationship data. Extracted temporal information is processing according to a set of temporal markers (heuristics) to determine a time value marker (i.e., an established time) of the security event. This processing typically involves retrieval of information from one or more structured data sources. The established time is linked to the security entities and relationships. The resulting security event, as augmented with the identified temporal data, is then subjected to a management operation.

Abstract: This disclosure provides an automatic causality tracking system that meets real-time analysis needs. It solves causality tracking for cybersecurity, preferably as three sub-tasks: backward tracking, forward tracking, and path-finding. Given a set of threat indicators, the first sub-task yields the system elements (e.g., entities such as processes, files, network sockets, and the like) that contribute information to a set of threat indicators backward in time. The second sub-task yields system elements forward in time. Given two sets of threat indicators, the third sub-task yields shortest paths between them, e.g., how the two sets of indicators are connected to one another. The system enables efficient multi-point traversal analysis with respect to a set of potential compromise points, and using data from real information flows.

Abstract: A cognitive security analytics platform is enhanced by providing a computationally- and storage-efficient data mining technique to improve the confidence and support for one or more hypotheses presented to a security analyst. The approach herein enables the security analyst to more readily validate a hypothesis and thereby corroborate threat assertions to identify the true causes of a security offense or alert. The data mining technique is entirely automated but involves an efficient search strategy that significantly reduces the number of data queries to be made against a data store of historical data. To this end, the algorithm makes use of maliciousness information attached to each hypothesis, and it uses a confidence schema to sequentially test indicators of a given hypothesis to generate a rank-ordered (by confidence) list of hypotheses to be presented for analysis and response by the security analyst.

Abstract: An automated method for processing security event data in association with a cybersecurity knowledge graph having nodes and edges. It begins by receiving from a security system (e.g., a SIEM) information representing an offense. An offense context graph is built. Thereafter, and to enhance the offense context graph, given nodes and edges of the knowledge graph are prioritized for traversal based on an encoding captured from a security analyst workflow. This prioritization is defined in a set of weights associated to the graph nodes and edges, and these weights may be derived using machine learning. The offense context graph is then refined by traversing the nodes and edges of the knowledge graph according to a prioritization tailored at least in part by the encoding. In addition to using security analyst workflow to augment generation of weights, preferably the machine learning system provides recommendations back to the security analysts to thereby influence their workflow.