Analytics Story: Suspicious Microsoft 365 Copilot Activities

Date: 2025-09-24 ID: 34cb1972-285e-4a3e-b235-d64246fcc8df Author: Rod Soto, Splunk Product: Splunk Enterprise Security

Description

Leverage advanced Splunk searches to detect and investigate suspicious activities targeting Microsoft 365 Copilot, including prompt injection attacks, agentic jailbreaks, information extraction attempts, compliance violations, and anomalous user behaviors.

Why it matters

Modern adversaries targeting AI systems employ increasingly sophisticated techniques that mirror traditional malware campaigns. Our detection framework identifies multi-stage attacks where threat actors use obfuscated prompts, layered social engineering, and persistent manipulation techniques to compromise AI security controls. These attacks often involve initial reconnaissance through seemingly benign requests, followed by escalated attempts to extract sensitive information or establish persistent behavioral modifications

Detections

Name ▲▼	Technique ▲▼	Type ▲▼
M365 Copilot Agentic Jailbreak Attack	Impair Defenses	Anomaly
M365 Copilot Application Usage Pattern Anomalies	Valid Accounts	Anomaly
M365 Copilot Failed Authentication Patterns	Brute Force	Anomaly
M365 Copilot Impersonation Jailbreak Attack	Impair Defenses	TTP
M365 Copilot Information Extraction Jailbreak Attack	Impair Defenses	TTP
M365 Copilot Jailbreak Attempts	Disable or Modify Tools	Anomaly
M365 Copilot Non Compliant Devices Accessing M365 Copilot	Impair Defenses	Anomaly
M365 Copilot Session Origin Anomalies	Valid Accounts	Anomaly

Data Sources

Name ▲▼	Platform ▲▼	Sourcetype ▲▼	Source ▲▼
M365 Copilot Graph API	N/A	`o365:graph:api`	`AuditLogs.SignIns`
M365 Exported eDiscovery Prompts	N/A	`csv`	`csv`

References

Source: GitHub | Version: 1