Xternal Studios
Back to home

Case study

Real-time cryptojacking detection and response

How a centralized metrics view correlated anomalous CPU, failing services, and intrusion signals in seconds—not hours.

GetBundled production dashboard: critical CPU, key services down, suspicious live process, and brute-force indicators.
GetBundled production dashboard: critical CPU, key services down, suspicious live process, and brute-force indicators.

1. The scenario (context)

A client runs a high-performance production server on a modern stack: Next.js, Docker, Redis, and SQL databases. Everything looked stable until the metrics UI raised a critical alert.

2. Detection: the GetBundled “aha” moment

From a single dashboard capture we identified three anomalies in seconds that plain-text logs could have taken hours to surface:

  • Process visibility

    The “Top processes (live)” view immediately showed a process named QsudaKR consuming 376.1% CPU.

  • Failure correlation

    The same screen showed that while CPU spiked, critical services (MySQL, PostgreSQL, Redis) were down.

  • Intrusion indicators

    The security panel showed 24 failed logins in 5 minutes—consistent with an active brute-force attempt.

Without a centralized visual interface, the client would have noticed slow web responses—but not the root cause: a malicious binary hiding among Next.js build artifacts.

3. Investigation (deep dive)

The metrics report supplied PID 738751 for a surgical response.

Malware path: the process ran from /app/.next/, pointing to web-layer risk such as RCE/SSRF exposure.

Resource impact: RAM at 50.4% and sustained CPU at 100.0% explained why the host sacrificed databases to stay up (memory pressure and OOM killer dynamics).

4. Resolution and hardening

  • Stopped the process for forensic analysis.
  • Identified the entry point (outdated Next.js and dependencies).
  • Removed malicious artifacts and upgraded dependencies.
  • Closed attack paths at the firewall, reducing login noise observed by the monitoring script.

5. Conclusion: the value of proactive monitoring

  • Lower MTTR: from hours to minutes.
  • Prevent larger damage: runaway cloud bills and customer data exposure.
  • Data-driven security: decisions from live process metrics, not guesses.

6. Efficiency: “invisible” monitoring vs invasive malware

A classic monitoring trap is the observer paradox: telemetry that costs so much CPU it distorts what it measures. In this incident the gap was stark.

  • Malware impact: QsudaKR was observed at 376.1% CPU and 29.7% RAM, starving legitimate services.

  • Tool impact: even under extreme load, our metrics reporter did not appear in the top 10 CPU consumers.

  • Visibility without penalty: at 100.0% CPU, collection kept flowing with a negligible footprint (under ~0.1% CPU), preserving signal when it mattered most.

The stack is designed to stay quiet: while malware fights for every cycle, the metrics path aims to put ground truth on your screen without taxing performance.