Stateful Runtime Detections Aim to Keep Pace With AI-Assisted Cloud Attacks

· Original article ↗

Summary

Sysdig argues stateful runtime detections can correlate container actions to reduce alert noise as cloud attacks accelerate, citing adoption and automation figures from its 2026 report and a Falco rule example.

Key points

  • Sysdig says attacks including React2Shell have been exploited within hours of disclosure; its Threat Research Team also reported AI-assisted attacks gaining admin access in eight minutes and stealing credentials in three.
  • Sysdig describes JADEPUFFER as the first agentic ransomware operation to autonomously run a destructive database extortion playbook.
  • Stateful detections correlate related events—such as opening a container shell, downloading a binary to /tmp, and executing it—to distinguish suspicious behavior from routine activity and reduce alert noise.
  • The article provides a Falco example using observations and linked process sessions to detect a package manager launched from a container shell.
  • Sysdig’s 2026 cloud-native security and usage report says 70% of organizations use stateful detections, covering 91% of cloud environments among adopters.
  • The same report says 75% of organizations have automated response actions configured, but only 27% have enabled them; Sysdig cites trust in detection quality as a barrier.
  • Sysdig reports a 140% increase in organizations using kill -9 after a threat detection alert and a 28% increase in organizations choosing to kill a container rather than pause or stop it.

Article Details

Defense Focus
Use stateful runtime detections to correlate cloud-native process activity, distinguish malicious behavior from routine operations, and support trusted, timely response.
Detection Methods
  • Correlate a terminal shell opening in a container with subsequent package-manager activity.
  • Detect the sequence of opening a shell, downloading a binary to /tmp, and executing it.
  • Link observations by shared process or session identifiers to build higher-confidence alerts.
Data Sources
  • Syscall telemetry, including execve events
  • Process and session identifiers
  • Container shell and process activity
Rule Types
  • Stateful Falco rules using observations and obs.occurs/obs.link conditions
Platforms
  • Falco agent
  • Sysdig Secure
Defensive Actions
  • Use correlated, higher-confidence detections to reduce false positives and investigation time.
  • Build trust in detection signals before enabling automated response actions.
  • For detected threats, organizations report killing processes or containers rather than pausing or stopping them.

MITRE ATT&CK

Vendors

Products

Tools

Related Articles