- endpoint-compromise
- malware
- persistence
- c2-communication
- manufacturing
- windows-7
Manufacturing Workstation Compromised: Multi-Stage Malware with Persistence and C2 Communication
Manufacturing workstation ws-001 shows strong multi-stage compromise evidence: malicious persistence installation via Startup folder, process injection into system processes, C2 communication via DGA domains and non-standard port 447, and re-execution of the same malicious binary 66 hours later. The attacker maintained a foothold despite CrowdStrike's quarantine actions.
Time and cost estimates are based on a Tier-2 SOC analyst model and actual investigation telemetry. Individual results vary by analyst experience, tooling, and environment.
Initial Signal
On 2026-04-26 at 10:05:53 UTC, CrowdStrike Falcon detected a multi-stage attack chain on ws-001, a Windows 7 manufacturing workstation at a Philippines facility. The process [CUSTOM_APP_1].exe spawned iexplore.exe instances that wrote a malicious executable (`kxnbrjvz.exe`, SHA256: [MALWARE_HASH_2]) to the user Startup folder—a known persistence mechanism—and attempted process injection into `iexplore.exe` and `dwm.exe`. The malware then queried DGA-characteristic domains (`xkrqnvbplt.com`, `wfsrtmnxqzpb.com`, `qhvnplxmrtw.com`, `bzrkvfmnxqplts.com`) and established an outbound connection to `[EXTERNAL_IP_2]:447` on a non-standard port, consistent with C2 communication.
What made this signal critical was not just the initial execution, but the persistence: the same malicious binary re-executed 66 hours later on 2026-04-29 from a different parent process (`explorer.exe`), indicating the attacker maintained control despite CrowdStrike's quarantine actions. The investigation correlated 14 CrowdStrike alerts across multiple MITRE ATT&CK tactics (Execution, Persistence, Defense Evasion, Command and Control) and 2,245 telemetry records over 3 days, revealing a persistent foothold and broader campaign activity on sibling hosts.
How We Reached the Verdict
The agent considered each plausible alternative verdict and ruled them out one at a time. Each card below lists the hypothesis tested, the evidence weighed, and the dismissal rationale.
Could this be a true positive that was successfully blocked?
Ruled outkxnbrjvz.exe quarantined by CrowdStrike on 2026-04-26iexplore.exe processes killed on 2026-04-26iexplore.exe killed on 2026-04-29kxnbrjvz.exe and killed injecting processes on 2026-04-26, the threat demonstrably persisted. The same malicious iexplore.exe binary (identical SHA256) re-executed 66 hours later on 2026-04-29 from a different parent process (explorer.exe instead of [CUSTOM_APP_1].exe), indicating the attacker maintained a foothold. The Startup folder persistence mechanism was installed before quarantine, and the 2026-04-27 Early Exploit Pivot Detect alert confirms continued unusual execution behavior between the two main events. A TRUE_POSITIVE_BLOCKED verdict requires that every identified activity was blocked with no successful unauthorized access — this condition is not met given the persistence installation and re-execution evidence.·High confidenceCould this be malicious execution without post-exploitation success?
Ruled outkxnbrjvz.exe)[EXTERNAL_IP_2]:447kxnbrjvz.exe) before quarantine; (2) C2 communication was attempted via DGA domains and non-standard port connections; (3) the malicious binary persisted and re-executed 66 hours later from a different execution path, indicating the attacker maintained access beyond the initial execution event. The COMPROMISED verdict is more appropriate when the attacker exercised control beyond initial intrusion.·High confidenceCould this be suspicious activity requiring further investigation?
Ruled outCould this be a false positive alert?
Ruled outDisconfirming Evidence
Evidence that pushed against the agent's working hypothesis. Each item changed the direction of the investigation.
Evidence Gathered
The agent queried these data sources during the investigation. Each entry shows what was checked, what came back, and the methodology behind the query.
kxnbrjvz.exe (SHA256: [MALWARE_HASH_2]) written by iexplore.exe to the user Startup folder (\Users\user_1\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\Startup\kxnbrjvz.exe) at 2026-04-26T10:05:55Z, establishing persistence. File was subsequently quarantined by CrowdStrike.[CUSTOM_APP_1].exe (C:\[CUSTOM_DIR_1]\[CUSTOM_DIR_2]\ws-001\[CUSTOM_APP_1].exe) spawned multiple iexplore.exe child processes at 10:05:53-10:05:56 UTC on 2026-04-26, which then performed malicious file writes and injection attempts. This parent-child relationship is anomalous for a manufacturing application.T1055) on 2026-04-26 at 10:06:57 UTC: iexplore.exe (PID 3888) attempted injection (blocked, process killed); iexplore.exe (PID 3888) second injection attempt (blocked); dwm.exe (PID 852) injection detected. Injection into dwm.exe is particularly significant as Desktop Window Manager is a high-value target for credential harvesting.xkrqnvbplt.com, wfsrtmnxqzpb.com, qhvnplxmrtw.com, bzrkvfmnxqplts.com — all exhibiting algorithmically generated naming patterns consistent with C2 domain generation algorithms. These queries occurred during the malware execution window on 2026-04-26.[EXTERNAL_IP_2]:447 at 2026-04-26T10:06:09Z. Port 447 is non-standard (not HTTPS/443), consistent with C2 communication on a non-standard port to evade detection.iexplore.exe instance (same malicious SHA256: [MALWARE_HASH_1]) launched from explorer.exe and loaded a malicious module (MaliciousModule alert, pattern 10136, T1129). Process was killed by prevention policy. The re-execution from a different parent (explorer.exe vs. [CUSTOM_APP_1].exe) and the same malicious binary hash indicates persistent presence of the threat.0.0.0.0 (all interfaces) across the entire 3-day observation window (7 NetworkListenIP4 events from 2026-04-26T05:10:48Z to 2026-04-29T04:30:41Z). Process attribution is unknown. Port 7777 has no standard service assignment; binding to all interfaces exposes it across all network segments. CrowdStrike classified this as Defense Evasion, Persistence, Command and Control.tvnserver.exe) actively receiving 18 inbound remote control connections during the observation window. TightVNC is a remote desktop/control application; its presence on an HMI/manufacturing workstation receiving active sessions is a significant remote access indicator.iexplore.exe (SHA256: [MALWARE_HASH_1]) appears in both the 2026-04-26 injection attempts and the 2026-04-29 module load event, confirming the same malicious binary persisted on the system across 66 hours despite quarantine actions.ws-002 and ws-003 have analyst-confirmed true-positive alerts involving [CUSTOM_APP_2].exe.exe (double .exe extension — known obfuscation technique) and [CUSTOM_APP_2]mgr.exe loading libraries matching known-malicious SHA256 hashes. PsExec (PSEXESVC.exe) matching Custom IOC hashes observed on multiple hosts (ws-004, ws-005), consistent with lateral movement tooling.HKCU\Software\Microsoft\Windows\CurrentVersion\Run) shows cyclical creation and deletion pattern across the observation window, classified by CrowdStrike as T1547.001 (Boot/Logon Autostart Execution) and T1070.009 (Indicator Removal). The deletion of internat.exe Run key entries (legacy persistence location) without corresponding creation events may indicate anti-forensic cleanup.False Positive Analysis
The agent ran these validation checks to confirm the verdict isn't a false positive.
- FP-01Evaluate whether the malicious file write (Fail
kxnbrjvz.exeto Startup folder) could be a legitimate software installation or update. - FP-02Evaluate whether the DGA-characteristic DNS queries could be legitimate application traffic.Fail
- FP-03Evaluate whether the 2026-04-29 MaliciousModule alert could be a false positive detection of a legitimate module.Fail
- FP-05Evaluate whether the broader campaign indicators (sibling host alerts, PsExec detections) corroborate or are independent of thePass
ws-001compromise.
Detection Opportunities
The high-fidelity detection signals from this investigation, sanitized for general use. Each MITRE ID links to the official technique reference.
| Technique | Tactic | Context |
|---|---|---|
T1547.001Registry Run Keys / Startup Folder | Persistence | Flag file writes to Startup folders (\Users\*\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\Startup\) by non-installer processes, especially iexplore.exe or other browsers. Alert on executable files with obfuscated or random names (e.g., kxnbrjvz.exe) written to persistence locations. Cross-reference file hashes against threat intelligence databases for known malware. The legitimate ways to deploy startup items go through MDM, GPO, or signed installers, not direct file writes from browsers. |
T1055Process Injection | Defense Evasion | Flag injection attempts from iexplore.exe or other user-mode processes into system processes (dwm.exe, svchost.exe, lsass.exe). Alert on multiple injection attempts from the same source process within a short time window (e.g., 2+ attempts in 60 seconds). Monitor for injection into Desktop Window Manager (dwm.exe) specifically, as it is a high-value target for credential harvesting and screen capture. Legitimate applications do not perform process injection; this is exclusively an attacker technique. |
T1568.002Dynamic Resolution | Command and Control | Flag DNS queries to domains with algorithmically generated naming patterns (high entropy, unusual character sequences, no recognizable words). Alert on queries to multiple DGA-characteristic domains from the same process within a short time window. Correlate DGA queries with outbound network connections on non-standard ports. Legitimate applications query well-known, human-readable domain names; DGA queries are exclusively associated with malware C2 communication. |
T1071.001Application Layer Protocol | Command and Control | Flag outbound TCP connections to external IPs on non-standard ports (not 80, 443, 22, 53, 123, etc.). Alert on connections to IPs with low prevalence or known malicious reputation. Correlate non-standard port connections with DGA DNS queries or process injection activity from the same source process. Port 447 is not a standard service port; connections to this port should be investigated immediately. Legitimate applications use well-known ports; non-standard ports are a common C2 evasion technique. |
T1129Shared Modules | Execution | Flag module loads by processes with known malicious hashes, especially when the same hash re-executes from different parent processes. Alert on module loads from non-standard directories (AppData, Temp, Downloads, user home directories). Monitor for module loads occurring hours or days after initial detection, as this indicates persistence. Cross-reference module hashes against threat intelligence. Legitimate modules are signed, versioned, and load from protected system directories; unsigned modules from user-writable locations are suspicious. |
T1571Non-Standard Port | Command and Control | Flag listening ports with no standard service assignment (port 7777, 8888, 9999, etc.) bound to 0.0.0.0 (all interfaces). Alert on persistent listeners across multiple days with unknown process attribution. Correlate unknown listeners with other compromise indicators (malware execution, injection, C2 queries). Legitimate services bind to specific interfaces and use well-known ports; binding to all interfaces on non-standard ports is a backdoor pattern. |
Verdict Reasoning
The verdict of Compromised at high confidence rests on the following mutually corroborating signals:
1. A pre-registered IOC hash match on the dropped file (kxnbrjvz.exe) written to a known persistence location (Startup folder), classified by CrowdStrike as known_malware
2. Multiple independent detection mechanisms firing in sequence: process injection into system processes (dwm.exe, iexplore.exe), DGA-characteristic DNS queries to four algorithmically generated domains, and outbound C2 connection to a non-standard port
3. Confirmed persistence: the same malicious binary (SHA256: [MALWARE_HASH_1]) re-executed 66 hours later from a different parent process, demonstrating the attacker maintained a foothold
4. Behavioral corroboration: eight Early Exploit Pivot Detect alerts clustered at the initial compromise time, plus one additional alert 32 hours later, confirming continued unusual execution behavior
5. Campaign-level evidence: sibling hosts (ws-002, ws-003) have analyst-confirmed true-positive alerts involving the same process names and IOC hashes, and PsExec activity across multiple hosts indicates lateral movement. Confidence is High rather than Confirmed because the process attribution for the persistent TCP 7777 listener remains unknown, and the TightVNC installation origin (attacker-deployed vs. pre-existing) cannot be definitively determined from available telemetry
Lessons
- 01Quarantine is not containment—verify persistence mechanisms. In this investigation, CrowdStrike quarantined the malicious executable on 2026-04-26, but the attacker had already installed persistence via the Startup folder before the quarantine action. The same malicious binary re-executed 66 hours later from a different parent process. Always audit what persistence mechanisms were installed before quarantine, not just what was blocked. Check Startup folders, Run keys, scheduled tasks, and services immediately after detecting malware execution. The blocked count is the distractor; the unblocked persistence is the real threat.
- 02Process parent-child anomalies warrant immediate escalation. A manufacturing application (
[CUSTOM_APP_1].exe) spawning multipleiexplore.exeinstances is anomalous and should trigger immediate investigation. In this case, the unusual parent-child relationship was the first signal of compromise. Establish baseline process execution patterns for critical applications in your environment. When a process spawns children that are inconsistent with its documented function, escalate immediately rather than waiting for additional alerts. This investigation took 8 minutes to complete; early escalation on parent-child anomalies can prevent persistence installation. - 03DGA queries + non-standard port connections = confirmed C2. This investigation detected four DGA-characteristic domains queried within minutes of an outbound connection to port 447. The combination of algorithmically generated domain names and non-standard port usage is a high-confidence C2 indicator. Do not wait for additional confirmation; treat DGA + non-standard port as confirmed command and control. Implement network-level detection for DGA patterns and non-standard port connections. The attacker in this case was attempting to establish C2 communication; early detection of this pattern could have prevented the 66-hour persistence window.
- 04Campaign-level corroboration reduces false positive risk. Sibling hosts (
ws-002,ws-003) had analyst-confirmed true-positive alerts with the same process names and IOC hashes. This corroboration across multiple independently monitored hosts significantly reduced the probability thatws-001detections were false positives. When investigating a single host, always check for related alerts on peer systems in the same environment. Shared IOC hashes, process names, or attack patterns across multiple hosts indicate a coordinated campaign rather than isolated incidents. This context strengthens confidence in the verdict and justifies immediate containment actions. - 05Windows 7 end-of-life systems are high-risk targets. This manufacturing workstation ran Windows 7 Professional (end-of-life January 2020), which no longer receives security patches. The attacker exploited this vulnerability gap to achieve persistence and C2 communication. Prioritize upgrading or isolating end-of-life systems, especially those in critical environments like manufacturing. If upgrade is not feasible, implement compensating controls: network segmentation, application whitelisting, and enhanced monitoring. The 66-hour persistence window in this investigation would have been much shorter on a patched, modern operating system.