Disclosure
🔥 CyberForge CVE of the Day #064
🚨 CVE-2026-101001 — Netcore NBR200V2 pre-authentication command injection
A permission-denied response can arrive too late. The researcher reports command execution before the router checks the requester's session. Investigate what the device did, as well as what its response said.
The VulDB CNA identifies OS command injection in Netcore NBR200V2 firmware 1.3.241127.071246, involving the web-management CGI's evaluation of query data. Its CVSS v3.1 base score is 10.0 Critical: network access, no prior privileges and no victim interaction.
⚠️ Restrict affected management interfaces now. No fixed firmware was verified in this review.
🎯 The quick hit:
Find the exact model and build. Establish who can reach management, including from internal networks. Preserve and review suspicious activity, then obtain a verified repair or supported replacement. The inspected vendor download identifies the affected build; it is not proof of a fix.
🔑 Key details:
⭐ Severity: Critical
📊 CVSS v3.1: 10.0 — VulDB CNA
🧮 Base vector: `CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H`
🧠 Weakness: CWE-78 — OS Command Injection; also CWE-77
🎯 Target: Netcore NBR200V2 1.3.241127.071246
🧩 Boundary: Request query data → shell evaluation in web management
🌐 Attack vector: Network
⚙️ Attack complexity: Low
🔓 Privileges required: None
👆 User interaction: None
🔀 CVSS v3.1 scope: Changed
⚔️ CNA-documented impact: OS command injection
💥 CVSS impact: High confidentiality, integrity and availability impact
👤 Execution context: Researcher reports root in firmware emulation
🛡️ Fix: No fixed build verified; restrict management reachability
🔑 Additional remediation: Rotate exposed secrets after trusted recovery where compromise warrants it
🚨 Active exploitation: Not confirmed by reviewed primary sources
🧪 Reproduction: Public researcher PoC; not executed by CyberForge
📋 KEV: No match in returned 2026.09.27 catalogue, checked 29 September
📉 EPSS: FIRST returned no data row
🏷️ CNA credit: llsss; original disclosure linked in the CVE record
Scoring note: CNA JSON supplies v4.0 10.0; NVD's CNA-sourced entry supplies 9.3 with `E:P`. Both are Critical. These are not independent assessments; the headline uses their consistent v3.1 base score.
🧬 What actually went wrong?
A diagnostic interface should keep incoming values as data and check permission before performing a sensitive action.
The disclosure describes an inactive input-processing call and a parser evaluating query data before checking the session: untrusted input reaches an interpreter before authorisation.
1️⃣ Input arrives
The CNA names `QUERY_STRING`, the CGI's request query data. This is not a requirement for a literal URL parameter with that name.
2️⃣ The parser interprets it
The researcher attributes the injection to shell evaluation in `/www/cgi-bin/network_tools`.
3️⃣ Permission checking follows
Their emulated-firmware test reports execution before a permission error is returned. That ordering informs the hunt; CyberForge has not reproduced it on a physical router.
No payload, weaponised request or execution recipe is included here.
A familiar component from Day
CVE-2026-94097 also names NBR200V2, the same build and the same CGI file. Its CNA description refers to `param/key/val`; today's record names `eval` and `QUERY_STRING` and links a different disclosure.
Both records remain published. We have not established whether they describe independent flaws or overlapping accounts of one root cause. Track both identifiers for affected assets, but do not infer a two-stage chain or count them as independently proven bugs.
⚔️ The practical attack chain:
1️⃣ A requester can reach the router's management service.
2️⃣ Query data reaches the vulnerable CGI handling path.
3️⃣ The reported parser passes untrusted input into shell evaluation.
4️⃣ Commands run under the affected service's execution context.
5️⃣ A later permission decision returns an error; any preceding side effects still require investigation.
The CNA supports the component and injection classification. Ordering and lab outcome are researcher-reported. Wider network access or data theft would require further evidence.
👤 Execution context — what the attacker actually gets.
The researcher reports root execution in emulated firmware. CyberForge did not verify a physical router's process identity or reproduce the test.
Scope the service's access to configuration, files and connected secrets. Traffic-policy control or credentials usable on another system would widen the investigation.
Persistence, interception, credential theft and lateral movement are potential consequences, not observed campaign facts in this edition.
📦 Affected products and versions:
🔴 Netcore NBR200V2
🔹 Explicitly affected firmware: 1.3.241127.071246
🔹 Component: Web Management Interface
🔹 Researcher-documented HTTP route: `/cgi-bin/network_tools`
🔹 Fixed firmware: not verified
The CNA's wildcard CPE does not expand its explicit affected-version list. Other builds and models are neither established as affected nor cleared by this record.
The vendor download identifies V1.3.241127.071246, the same affected version with a leading V. Ask for a fix tied to the CVE, model and hardware revision.
🕰️ Timeline / exploitation window.
27 September: CNA-listed advisory disclosure. 28 September: CVE publication and CISA ADP's poc assessment. 29 September: this review.
No earliest real-world attack date is established. Set the hunt window from exposure history and retention; publication does not clear earlier activity.
👁️ Defender hunting guide:
1️⃣ Establish the exposure window
Record model, firmware, owner, management listeners, approved administrator sources and the time restrictions took effect. Include WAN, client and guest networks, VPN access, remote-support paths and IPv6. A login page confirms a reachable service, not a safe pre-authentication boundary.
Use configuration history to assess previous access; today's rules may differ.
2️⃣ Review management requests
Where existing logs provide visibility, review the documented CGI route for unusual sources, query structure, bursts or activity outside maintenance periods. Compare with legitimate diagnostic use and approved automation.
Keep permission errors in scope. The reported sequence makes request-to-side-effect correlation more useful than a filter that keeps only apparent successes. Neither an HTTP error nor an application error settles the execution question.
Check what the router actually logs. Preserve raw evidence alongside normalised copies, document gaps and protect sensitive request data.
3️⃣ Review device activity
Use existing telemetry or an approved evidence-collection process to look for unexpected commands, writes, new processes or configuration changes around suspicious request times.
The CGI is itself a shell script. A shell beneath the web service is therefore not sufficient evidence of compromise. Focus on activity inconsistent with the requested diagnostic task, then verify against known administration work.
Without process auditing, use configuration backups and upstream records. Missing telemetry cannot clear the device.
4️⃣ Correlate appliance egress
Review DNS, firewall and flow records for new destinations or unusual outbound sessions originating from the appliance. Align clocks and join activity with management events.
Baseline DNS, NTP and maintenance traffic before treating new destinations as suspicious. No campaign IP, domain or malware hash was established here.
5️⃣ Scope reachable identities and settings
Compare administrator accounts, remote-access settings, DNS servers, routes, forwarding rules and startup configuration against a trusted baseline. If evidence supports compromise, determine which administration or VPN secrets were accessible.
Record evidence for each change and keep credential-compromise claims scoped.
🩹 Emergency remediation order:
1️⃣ Restrict management reachability
Use trusted upstream controls where possible. Allow necessary administration paths and assess the effect of isolation on business traffic.
2️⃣ Preserve evidence where safe
Retain logs, flow data, configuration history and maintenance records. Use approved collection methods and do not delay urgent containment.
3️⃣ Obtain a verified remedy
Seek a vendor-confirmed fix for the exact hardware. Otherwise plan replacement or removal from the affected role. An upgrade label does not prove this CVE was addressed.
4️⃣ Verify what changed
Record image provenance, repair evidence and the running build. Verify ordinary access from approved administration systems and denial from authorised test locations in disallowed networks. These are access-control checks, not exploit tests.
5️⃣ Recover trust and rotate relevant secrets
For confirmed or unbounded compromise, recover or replace through a trusted process. Rotate exposed credentials after containment and scope dependent systems from evidence.
6️⃣ Close with evidence
Record repair verification, hunt coverage and the monitoring owner. A reboot or scheduled upgrade is insufficient closure evidence.
🧱 Temporary exposure reduction.
✅ Restrict management at a separate trusted network boundary where feasible.
✅ Include IPv4, IPv6, alternate listeners and internal access paths in the review.
✅ Limit approved administrator sources; ordinary client and guest access is unnecessary exposure.
✅ Review appliance egress and send available logs to a trusted collector.
✅ Give temporary restrictions an owner and a date for repair or replacement review.
Changing the administrator password alone cannot address processing before authentication. Ad hoc input filtering or editing a CGI is not a verified vendor repair. Containment reduces opportunity; it does not establish that earlier compromise is absent.
🚑 When vulnerability management becomes incident response.
Escalate suspicious management requests correlated with unexplained device activity, unauthorised configuration changes, new accounts or unusual outbound connections. Assess evidence gaps alongside exposure and the asset's role; an exposed version alone does not prove compromise.
Contain through trusted infrastructure, preserve records and scope credentials and connected services. If the appliance's integrity cannot be established, replacement or trusted recovery may be more defensible than deleting one suspicious file.
📊 CISA KEV and EPSS context.
Checked 29 September 2026: no exact match in the returned KEV catalogue 2026.09.27. Its release predates this CVE's publication, so this is a dated catalogue result, not assurance about current attack activity.
The later CISA ADP assessment says poc, automatable yes, technical impact total. ADP classification and KEV membership answer different questions.
FIRST's direct API query returned no EPSS row. An unreviewed mirror carried a number, but this edition does not adopt it without direct confirmation and a score date. Missing data is not a probability of zero. Prioritise the documented impact and your actual exposure.
🧾 Evidence separation.
CNA-supported: named device/build, CGI input and function, command-injection classification, critical scoring and public exploit reference.
Vendor-supported: the inspected firmware page lists the affected build. A security fix was not verified.
Researcher-reported: execution before the session check, the HTTP route and root execution in emulated firmware.
CyberForge interpretation: prioritise management reachability, correlate errors with side effects, and verify repair or replacement before closure.
Not established: a field campaign, physical-device reproduction by CyberForge, independent root causes versus , a fixed build or a complete affected-version range.
🔥 CyberForge verdict:
CVE-2026-101001 combines critical management-interface injection, public technical detail and an unresolved repair path. The timing of a permission check changes what defenders can infer from a response.
The correct order is:
1️⃣ Restrict the reachable management surface.
2️⃣ Preserve and review prior exposure evidence.
3️⃣ Obtain and verify a supported repair or replacement.
4️⃣ Resolve compromise and exposed credentials where warranted.
5️⃣ Monitor the recovered device and relevant connected systems.
> A permission error is not proof of prevention when the dangerous work happened first.
Research cutoff: 29 September 2026, 09:21 BST / 08:21 UTC. Source review and document validation only; no exploit or router test was performed.
🔗 CNA / CVE record:
https://www.cve.org/CVERecord?id=CVE-2026-101001
🔗 NVD:
https://nvd.nist.gov/vuln/detail/CVE-2026-101001
🔗 Vendor firmware page — affected build, not a verified fix:
https://www.netcoretec.com/service-support/download/firmware/2705.html
🔗 Original researcher disclosure — includes offensive material; not reproduced here:
https://github.com/senxitoyshuyi-ui/HACKALL/blob/main/netcore_NBR200V2_V1.3.241127.071246%20Router/Netcore_NBR200V2_network_tools_command_injection.md
🔗 Related record / Day :
https://www.cve.org/CVERecord?id=CVE-2026-94097
🔗 CISA KEV:
https://www.cisa.gov/known-exploited-vulnerabilities-catalog
🔗 FIRST EPSS query:
https://api.first.org/data/v1/epss?cve=CVE-2026-101001