How Darktrace Used Anomaly-Based Threat Hunting to Investigate a Suspected Energy-Sector Attack

Summary
Darktrace describes using experimental detection models to hunt for PerfectData-related activity, uncovering a suspected attack involving email access, deletions and downloads of operational files at an energy-sector organization.
Key points
- Darktrace describes anomaly-based models that characterize normal device and user behavior and let security teams create or adjust detections.
- Researchers built an experimental model to look for SaaS activity associated with PerfectData, alongside existing SaaS and email detections.
- The hunt surfaced alerts involving PerfectData use at energy-sector customers, including an oil and gas investment company.
- In a reported June 2024 case, an actor accessed Azure AD, granted application permissions, created an email rule, accessed sent mail and mass-deleted messages that may have been used in phishing.
- The same actor later downloaded mLog CSV files and operational documents, which Darktrace said suggested possible interest in OT attack vectors.
- Darktrace said it notified the affected customer after verifying its findings, and the customer’s security team took steps to prevent further malicious activity.
Article Details
- Defense Focus
- Proactively hunt for anomalous activity associated with a hypothesis-driven attack, using experimental and existing detection models to identify suspicious SaaS, email, and file activity.
- Detection Methods
- Establish a behavioral baseline using machine-learning analysis of device and actor activity, then investigate anomalous behavior unlikely to be benign.
- Create experimental models for specific hypotheses, such as SaaS activity involving PerfectData Software, and analyze the resulting alerts.
- Correlate experimental alerts with existing SaaS- and email-related model alerts to investigate activity across multiple stages.
- Review alerts for unusual account sign-ins, application consent and service-principal activity, email-rule changes, mailbox access and deletion, and bulk file downloads.
- Use the Model Editor to configure detection values, priorities, thresholds, and actions, and adjust existing detection priorities based on environmental risk.
- Data Sources
- Darktrace Deep Packet Inspection outputs and telemetry inputs
- SaaS activity and related detection alerts
- Email activity and related detection alerts
- Account sign-in, application-consent, and service-principal activity
- Email-rule changes, mailbox access and deletions
- File-download activity, including mLog CSV files and operational files
- Rule Types
- Custom experimental anomaly-detection models
- Existing SaaS- and email-related detection models
- Platforms
- Darktrace
- Defensive Actions
- Run hypothesis-driven hunts and test experimental models against relevant activity.
- Correlate anomalous alerts with existing detections and investigate the activity before treating it as malicious.
- Tune model thresholds, priorities, and actions to fit use cases and organizational risk.
- Notify the impacted customer after findings are verified and have the internal security team take steps to prevent further malicious activity.
MITRE ATT&CK
T1070.008 · Clear Mailbox DataThe article describes the actor mass-deleting sent emails, apparently to obscure a potential outbound phishing campaign.T1114.003 · Email Forwarding RuleThe actor created an email rule that moved messages to an RSS Feed Folder, stopped processing other rules, and marked messages as read.
People
Vendors
Products
Azure ADThe actor logged into Azure AD from a rare US IP address. They then granted Consent to ‘eM Client’ from the same IP. Shortly after, the actor granted ‘AddServicePrincipal’ via Azure to PerfectData Software. Two dayseM ClientThe actor logged into Azure AD from a rare US IP address. They then granted Consent to ‘eM Client’ from the same IP. Shortly after, the actor granted ‘AddServicePrincipal’ via Azure to PerfectData Software. Two daysModel Editordetection, there is also the ability to change and create models showcasing the tool’s diversity. The Model Editor allows security teams to specify values, priorities, thresholds, and actions they want to detect. ThatPerfectData SoftwareFor one of the hunts, which hypothesised utilization of PerfectData software and multi-factor authentication (MFA) bypass to compromise user accounts and destruct data, an experimental model was created to detect aPrivate Internet Access Virtual Private Network (PIA VPN)in the “\Sent” folder from a malicious IP belonging to anonymization network, Private Internet Access Virtual Private Network (PIA VPN) [1]. The actor then conducted mass email deletions, deleting multiple instances
Countries
Industries
EnergyConducting a threat hunt in the energy sector with experimental modelsOil and gasmodel associated with PerfectData software usage, within energy sector customers, including an oil and gas investment company, as well as other sectors. Upon further investigation, it was also found that inRenewable energyUpon further investigation, it was also found that in June 2024, a malicious actor had targeted a renewable energy infrastructure provider via a PerfectData Software attack and demonstrated intent to conduct an
