Security Vision UEBA automatically builds typical models of behaviour (users, accounts, devices, processes, etc.) and finds outliers by analysing raw data streams of network traffic, proxy servers, mail servers, Windows/Linux servers and workstations.
Security Vision Anomaly Detection extends the ability to detect anomalies in the corporate infrastructure by applying a large number of different Machine Learning models and techniques, stacking the results of individual models and aggregating the resulting events into incidents for further investigation.
The most significant capabilities of Security Vision UEBA:
Integration with data sources
The Security Vision UEBA product contains configured connectors for receiving, normalising and analysing raw data from all popular SIEM systems (KUMA SIEM, MaxPatrol SIEM, Pangeo RADAR, RuSIEM, NEURODAT SIEM, ArcSight SIEM, QRadar, Splunk, etc.), the ability to receive events in universal formats (CEF, LEEF, etc.), connectors to NG systems (KUMA SIEM, MaxPatrol SIEM, Pangeo RADAR, RuSIEM, NEURODAT SIEM, ArcSight SIEM, QRadar, Splunk, etc.), the ability to receive events in universal formats (CEF, LEEF, etc.), and connectors to NG systems. ), connectors to NGFW and network devices (Cisco, CheckPoint, PaloAlto, Juniper, etc.), proxy servers (Squid, Blue Coat), ‘data lakes’ (Kafka, Elasticsearch), as well as receiving logs directly from Windows/Linux devices and workstations.
Integration builders built into the platform allow you to quickly implement additional integrations with any other data sources using a large number of protocols in no-code mode, including a graphical builder for normalisation of the received data.
Customisable rules and analytics engine
Users have access to dozens of built-in rules for statistical analysis of various parameters of user activity, accounts, hosts, processes, as well as traffic volumes, number of connections, etc. The product's functionality allows flexible expansion and customisation of new analysis rules, setting their activity, evaluation and threshold of influence on the creation of the final incident.
The platform also includes a full-fledged correlation rules engine, which can be used to configure rules of any depth and complexity. For example, the product comes with sigma rules and typical correlation rules.
Incidents and response
All detected deviations are automatically aggregated with respect to the alarm object. When specified thresholds are exceeded, the system generates an incident that reflects all detailed information about the incident object, related objects and all detected abnormal events.
Automated actions are configured in the product to process incidents: sending to IRP/SOAR systems, sending to SIEM, adding to SIEM Active List, adding to blocking lists on NGFV, blocking in directory service, etc. The user can adjust the actions performed by the product by settings. The user can adjust the actions performed by settings: enable their execution automatically or manually, control their visibility on the incident card. Incident notifications can be managed in the same way.
The system automatically creates separate objects for all related incident attributes (devices, accounts, etc.). For each object, collection and enrichment with additional data from the customer's infrastructure or from external analytics services are automatically triggered. The process of data collection and enrichment is regulated by the system settings.
To work with identified incidents and related objects, the product has built-in workflows that manage the incident lifecycle, enrichments, and enable actions. The workflow builder built into the platform allows users to customise the required response process and configure interactions with external systems.
Flexible options are available in the platform to create and customise additional response actions, data collection and enrichment of both the received incident and all associated customer infrastructure or external systems.
Visualisation and reporting
In the incident card, all detected events for an object are displayed in Timeline view, following the chronology of their occurrence. A large number of links to related objects (devices, accounts, processes, etc.) allows you to navigate to their cards for additional data and analysis.
Additionally, all related objects and attributes are displayed in the form of a graph, which allows you to build links between incident objects and quickly jump to detailed information on them. The user can add additional actions to the graph to respond, enrich the data, or build additional relationships.
General views and dashboards allow you to view summary information on all identified objects, according to a calculated rating. Drill-down allows you to view detailed information for each analysis group.
For each object in the system there is a possibility to upload reports containing all detailed information on detected alarms and objects of violations. Summary reports for a period can be uploaded manually or received on schedule via various channels: e-mail, Telegram, etc.
The reporting and dashboard builder built into the platform allows users to independently configure the required reporting and data visualisation in no-code mode without the use of any external products and tools.
The most significant capabilities of Security Vision Anomaly Detection:
In addition to all of the above capabilities, the user receives a large number of preconfigured and trained Machine Learning models that significantly enhance the ability to detect anomalous and suspicious activities in the corporate infrastructure that are not detected by correlation rules and functionality of standard antimalware systems.
The product uses various ML techniques, trained models on various datasets related to the activity of botnets, VPO, DDOS attacks, etc., ‘teacherless’ models that automatically approximate activity and detect deviations by various combinations of parameters, neural networks that take into account the sequence of events and their relationships, etc. The resulting alarms are automatically processed, grouped into sets of events, and data deduplication is applied.
The models used in the product are automatically retrained on the customer's data on a regular basis, adapting to the infrastructure settings, network, technical and user activity. Both manual and automatic selection of model parameters is used to improve the quality of detected alarms.
The product has built-in whitelisting capabilities to customise exceptions. Also, the models automatically take into account false-positive alarms during the subsequent retraining of the models.