Cybersecurity

Preventing alert fatigue in Cortex XDR

Security teams face significant alert fatigue, with a Dimensional Research survey showing 83% of staff struggle with notification noise. Cortex XDR helps mitigate this through granular severity controls and SmartScore, which uses machine learning to provide adaptive risk scoring.

Preventing alert fatigue in Cortex XDR

Security teams drown in low-priority noise, burying critical detections under mountains of notifications. A Dimensional Research survey shows that 83% of security staff face alert fatigue, and 93% of respondents cannot address all security alerts on the same day. SOC teams want to minimize this noise to focus limited resources on necessary tasks, especially because every environment has unique risk tolerance, environment specifics, and organizational policies. If an analyst tunes a behavioral rule to a high severity for common, low-risk activity, they create an unmanageable alert flood. Setting severity to an informational or low level can cause a team to lose sight of genuine threats. Cortex XDR includes impact warnings to prevent these mistakes. These prompts appear when users set a rule’s severity to low, informational, high, or critical levels. The tool allows granular control over alert severity, including critical, high, medium, low, and informational levels. Users can achieve optimal triage by adjusting severity to match their risk profile. A threat against a sensitive database requires immediate critical escalation, while other assets carry less business risk. Teams use the reversion option to change custom settings back to default values.

An analyst might encounter a "Rare Remote Service Command Execution" rule that flags high severity in most environments. In environments that run frequent, automated, and benign remote scripts for IT configuration management, such as nightly patch installations managed by a unique automation service account, this rule generates consistent, expected, and high-volume alerts. The SOC team can change the severity for this specific rule to medium or low when it triggers from that account. This customization reduces alert noise in the high-priority queue, saving analyst time without sacrificing the underlying detection logic. If you are managing a multi-tenant environment, role-based access control (RBAC) limits modifications to authorized users. The feature supports multi-tenant managed environments, which helps MSSP use cases. The feature integrates role-based access control (RBAC) to limit modifications to authorized users. Unauthorized users or those without necessary expertise risk creating a significant security gap if they modify global security rules. You must ensure that modifications stay within authorized hands to avoid widespread operational outages. Does reducing severity for automation accounts create a gap for attackers to hide?

Traditional scoring mechanisms are often static and time-consuming. These methods miss the context required to reflect the real score of a threat. Incidents can reoccur with the same score even if analysts previously marked them as false positives in the past. SmartScore addresses this by using machine learning and behavioral analysis to generate a risk score for each incident. It uses an ensemble of Gradient Boosting models to generate a unified risk score. The engine uses context-driven features, incident-level features, statistical features, and model-based features. Context-driven features rely on customer resolutions and manual research conducted by experts. Incident-level features quantify threat statistics like the number of alerts, their severities, and where the alerts originated. Statistical features leverage the Cortex Data Lake to measure the prevalence of entities like files, domains, and IPs in a tenant-based and global manner. Model-based features use independent ML models to monitor specific aspects, such as the CMD process tree, to assess risk. If a team uses both custom incident scoring and SmartScore, custom scores override the scores from the SmartScore engine. This scoring engine is adaptive to changes in the environment and to changes in alert behaviors. SmartScore translates security investigation methods and associated data into a hybrid scoring system. This score helps with incident triaging and threat hunting. SmartScore predicts risk levels based on cyber-oriented features.