Appearance
Insider Risk & IOC Detection
GhostAI's Insider Risk system provides two complementary layers of protection: rule-based Indicator of Compromise (IOC) detection and statistical behavioral anomaly detection. IOC rules evaluate audit events in real time and in periodic batch scans against administrator-defined patterns. Behavioral baselines learn normal user activity from historical audit data and flag statistical deviations that may indicate account compromise, misuse, or insider threats.
How It Works
The architecture combines deterministic rules with statistical baselines:
IOC Engine — Six rule types (
volume_spike,off_hours,geo_anomaly,data_exfil,access_pattern,custom) are evaluated against every relevant audit event in real time. The same definitions are also applied during scheduled batch evaluation over rolling time windows for coverage and backfill.Behavioral anomaly detection — Per-user baselines are computed from audit activity; current metrics are compared using z-score deviation so that spikes beyond normal variance generate findings without relying on fixed static thresholds alone.
IOC Rule Types
| Type | Description | Pattern fields |
|---|---|---|
volume_spike | Unusual query or activity volume relative to the user’s baseline | metric, threshold_sigma, min_count |
off_hours | Activity outside configured business hours | start_hour, end_hour, timezone, min_events |
geo_anomaly | Access from blocked countries or regions, or excessive geographic diversity (country tier system for blocked lists) | blocked_countries, max_countries_per_day |
data_exfil | Elevated download or export patterns suggestive of bulk movement of data | metric, threshold_sigma, min_count |
access_pattern | Suspicious frequency of selected audit event types (from the event catalog) | resource_types, max_per_hour |
custom | Flexible matching using regex and optional structured filters | event_type_regex, detail_match (JSON) |
Severity Levels
| Severity | When to use |
|---|---|
| low | Informational detections, tuning, or early rollout where you want visibility without paging on-call |
| medium | Typical production rules that warrant review but are not assumed to be incidents by default |
| high | Strong signals that should be triaged promptly and may correlate with other security data |
| critical | Severe or time-sensitive conditions; reserve for patterns with very high confidence or regulatory impact |
Severity affects alert prioritization, how matches surface in history, and the severity carried on integrated SIEM events.
Alert Channels
SIEM event — When a rule matches or a behavioral anomaly is detected, GhostAI can emit structured events such as
insider_risk.ioc_matched(and related insider-risk event types) into the centralized SIEM integration pipeline. Configure the delivery endpoint under Admin → Security → SIEM Integration.Email notification — Rules can notify organization administrators and any additional recipient addresses configured on the rule, so the right owners receive matches outside the SIEM when needed.
Behavioral Baselines
Baselines are computed from audit log activity over a configurable lookback window (commonly on the order of 7–90 days depending on organization settings and API parameters). Metrics include query and activity volume, session-oriented signals, distinct IP or geographic indicators, file download patterns, and related aggregates. Anomalies are raised when current activity deviates from the learned distribution beyond configurable z-score thresholds; typical defaults are warning at 2.5σ and critical at 3.5σ for scheduled scans, with on-demand scans respecting the thresholds you configure in Detection Settings.
Configuration
IOC rules, match history, baselines, and detection behavior are managed in the Admin Dashboard under Insider Risk. The area is organized into four sub-tabs: IOC Rules, Match History, Baselines, and Detection Settings.
TIP
Navigate to Admin Dashboard → Insider Risk to configure IOC rules, review match history, manage behavioral baselines, and adjust detection settings.