Volt Typhoon's Silent Train: Four Zero-Days in Management Tools
On October 28, 2026, Lumen's Black Lotus Labs published research on Silent Train. It is a covert access framework the lab attributes to the China-nexus group Volt Typhoon. The report ties the framework to four confirmed zero-day campaigns against centralized management platforms: SD-WAN controllers, endpoint management and network control planes. This article separates what is confirmed from what public reporting does not yet establish.
What happened
Black Lotus Labs says it has tracked Silent Train for about three years. Per Lumen, the framework was first identified in late 2023, when researchers noticed unusual SSH relay chains while analyzing a separate covert network. Lumen describes Silent Train as rotating chains of compromised SOHO routers, IoT devices and enterprise infrastructure. It says the operators use anti-forensic techniques, compartmentalized operations and post-operation cleanup.
Lumen says its visibility into the framework let it send hundreds of victim notifications to public and private sector partners. Some of those notifications led to incident response investigations. Those investigations uncovered zero-day exploits used for initial access. The Hacker News, BleepingComputer, CyberScoop and Nextgov all covered the findings on the day of publication.
Researchers say Volt Typhoon leaned on Silent Train more heavily after the FBI and Justice Department disrupted the KV-botnet in early 2024. Lumen describes the framework as "now-disrupted" or quietly disrupted. Because the infrastructure is rebuilt constantly, though, a researcher told CyberScoop that what exists today will be completely different tomorrow. Treat the threat as ongoing even if specific chains are gone.
The four confirmed zero-day campaigns
- Versa Director — CVE-2024-39717, a file upload flaw. It was previously reported in August 2024 with delivery of the VersaMem web shell, which harvests credentials and gives access to downstream customers.
- N-able N-central — CVE-2025-8875 and CVE-2025-8876, insecure deserialization and command injection.
- Cisco Catalyst SD-WAN Controller — CVE-2026-20127, an authentication bypass.
- Arista VeloCloud Orchestrator — CVE-2026-16812, an OS command injection flaw.
Lumen also reports one likely zero-day in Quest KACE, patched in October 2025. That one is a suspicion, not a confirmed finding.
Who is affected
Victims. Reporting says hundreds of victims were identified. Sectors include telecommunications providers, government, defense contractors, satellite and mission-critical communications, universities, IT firms and MSPs. Critical infrastructure victims span water, energy, transportation, maritime, rail, aviation and emergency services. Public reporting does not give an exact victim count, a per-sector breakdown or a list of named organizations.
Geography. Lumen says Hawaii, California and Texas repeatedly emerged as focal points. The researchers read this as consistent with positioning against defense and infrastructure networks tied to the Indo-Pacific. That is an assessment of intent. Lumen's findings do not show that any disruptive action has occurred.
Technology. Lumen says the group exploited or abused more than 27 device classes. Vendors named as compromised include Fortinet, Cisco, ManageEngine, Ivanti, F5, Check Point, Palo Alto Networks, ServiceNow, TP-Link, Netgear and ArcGIS. Many of these were reportedly hit through known, patchable vulnerabilities, not zero-days. Lumen also noted that some compromised devices had no known flaws, which suggests the actors may hold other undiscovered vulnerabilities.
Who should worry most. The common thread is management planes that hold trust over many other systems:
- SD-WAN orchestrators and controllers
- RMM and endpoint management tools
- Identity and access management
- Out-of-band management platforms
Lumen says the actors favored entities that could provide scalable access to downstream customers. MSPs, ISPs and telecoms are therefore at elevated risk.
Attribution caveat. Lumen attributes Silent Train to Volt Typhoon. Public reporting reviewed here does not state a formal confidence level or any government attribution specific to this report. The group is also tracked as Bronze Silhouette, Insidious Taurus, UNC3236, Vanguard Panda and Voltzite.
How the attack works
Silent Train is less a botnet than an operational workflow. Reporting describes these stages:
- Relay chain assembly. Operators route traffic through VPS infrastructure outside the United States. They then pass it through hacked routers and other devices acting as intermediate hops. BleepingComputer reports Intelbras modems served as overseas relays, and Netgear Orbi devices served as U.S. hops.
- Disposable build. Lumen says the operators used temporary SSH services, keys, configuration files and living-off-the-land utilities on the relays. A CyberScoop quote from a Lumen researcher says a new chain with different infrastructure can be built in under two minutes.
- Exploitation. The chain delivers exploits against management platforms, including the four zero-days above.
- Post-exploitation. Lumen cites in-memory web shells, credential harvesting and abuse of native features such as endpoint control and remote file transfer. In several cases, the actors erased or altered logs and other evidence.
- Teardown. Sessions typically end in under eight hours (an average of roughly four hours is also reported). The operators then remove tooling, keys and configuration files from the relays. Researchers say this left "no evidence for defenders to uncover" on the relay side.
This design makes the usual IOC-based approach weak. A relay IP is likely to be a compromised consumer or branch device that is clean a few hours later. The durable evidence is on the victim's management platform and on the network edge facing it.
MITRE ATT&CK mapping
These mappings are the analyst's reading of the reported behavior, not a technique list published by Lumen.
- T1190 — Exploit Public-Facing Application (zero-day exploitation of management platforms)
- T1584.008 — Compromise Infrastructure: Network Devices (hijacked SOHO/IoT relays)
- T1583.003 — Acquire Infrastructure: Virtual Private Server (offshore VPS entry points)
- T1090.003 — Proxy: Multi-hop Proxy (rotating relay chains)
- T1021.004 — Remote Services: SSH (temporary SSH services on relays)
- T1505.003 — Server Software Component: Web Shell (in-memory web shells such as VersaMem)
- T1070 — Indicator Removal (log erasure and artifact cleanup)
CVEs and IOCs
The CVEs are listed above: CVE-2024-39717, CVE-2025-8875, CVE-2025-8876, CVE-2026-20127 and CVE-2026-16812. This article does not reproduce a list of IP addresses, hashes or domains. Lumen's technical report is the authoritative source for any published indicators. Given the ephemeral design, expect indicators to age quickly.
One caution on Arista: other VeloCloud Orchestrator advisories and exploited bugs have appeared in recent weeks, some with different CVE IDs. Public reporting reviewed here does not tie them to Volt Typhoon. Match advisories to CVE-2026-16812 rather than assuming every VCO fix covers it.
Detection guidance for defenders
Assume relay-side evidence is gone. Focus on the management platform itself, the edge devices around it, and who has been talking to both.
- Management plane access. Look for admin or API logins to SD-WAN controllers, N-central, KACE and similar tools from residential ISP, hosting or unfamiliar ranges. Also flag sessions that start and stop within a few hours.
- Web server child processes. Look for shells, interpreters or downloaders spawned by the management application's service account (T1190, T1505.003).
- Java app anomalies. Look for unexpected classes loaded into the web application's JVM, and for web server processes that read credential stores. These are signs of in-memory web shells.
- Log tampering. Look for gaps, truncation or deletion in application, auth and syslog files on appliances (T1070). A sudden log gap on a management host is itself an alert.
- New SSH artifacts. Look for new
authorized_keysentries, new sshd listeners on odd ports, or new local accounts on edge devices. - Credential reuse. Look for service accounts or API tokens from the management platform authenticating to downstream devices or customer tenants from new sources.
Example hunting query (KQL, adapt to your schema)
// Inbound connections to management-plane hosts from non-allowlisted sources,
// ranked by how few distinct days they appear (short-lived, one-off sources)
let MgmtHosts = dynamic(["sdwan-controller01","ncentral01","vco01","kace01"]);
let Allowlist = externaldata(IP:string)[@"https://your-storage/mgmt_allowlist.csv"] with (format="csv");
DeviceNetworkEvents
| where Timestamp > ago(30d)
| where DeviceName in~ (MgmtHosts)
| where ActionType == "InboundConnectionAccepted"
| where LocalPort in (22, 443, 8443)
| where RemoteIP !in (Allowlist)
| summarize Sessions=count(), Days=dcount(bin(Timestamp,1d)),
First=min(Timestamp), Last=max(Timestamp) by DeviceName, RemoteIP, LocalPort
| where Days <= 2
| order by Sessions desc
What it finds: one-off or short-lived external sources reaching management platforms. This is the pattern a rotating relay chain leaves behind. Enrich the results with ASN data and prioritize consumer ISP and hosting ranges.
Second hunt (generic pseudo-query for appliance and Linux logs)
source=syslog OR auditd host IN (management_platforms, edge_devices)
(event="file_delete" OR event="file_truncate")
path IN ("/var/log/*", "*/authorized_keys", "*/.ssh/*")
| stats count by host, user, path, process
| where user != "root" OR process NOT IN (logrotate_allowlist)
What it finds: log or SSH key deletion outside normal rotation, which can indicate post-operation cleanup.
Retention matters. Pull the longest history you have for management platform auth logs, reverse proxy logs and NetFlow/firewall logs. Hunt back at least to the earliest vulnerable window for each CVE. N-central and Versa exposure dates reach back to 2024 and 2025.
Mitigation and response
- Inventory management platforms first. Find every instance of Versa Director, N-able N-central, Cisco Catalyst SD-WAN controllers and VeloCloud Orchestrator (on-prem and self-hosted). Note which are internet-reachable.
- Patch to the vendor-fixed releases for the CVEs above. Use each vendor's advisory for the exact fixed version (Versa for CVE-2024-39717, N-able for CVE-2025-8875/8876, Cisco for CVE-2026-20127, Arista for CVE-2026-16812). This article does not list fixed build numbers, because reporting reviewed here does not include them. Also check Quest KACE is on a release patched in or after October 2025.
- Take management interfaces off the internet. Restrict them to a jump host, VPN or allowlisted admin ranges. Zero-days in these products will keep appearing, so exposure reduction outlasts any one patch.
- Treat patching as not equal to clean. Researchers say the actors cleaned up after themselves. If a platform was exposed and unpatched in the relevant window, run a compromise assessment. Preserve disk and memory images before rebooting, since in-memory implants vanish on restart.
- Rotate credentials and secrets. If compromise is plausible, rotate admin passwords, API tokens, SSH keys, certificates and any credentials stored on or managed by the platform. Include downstream customer or tenant credentials for MSPs.
- Review downstream trust. Check what the management platform can push or execute on managed endpoints, branch devices or customer networks. Look for unexpected policy, template or script changes.
- Harden the edge. Replace end-of-life SOHO and branch routers that no longer receive patches. Enforce MFA and out-of-band logging for appliance administration, and ship logs off-box in real time so local deletion does not erase evidence.
- Report and coordinate. U.S. organizations in the affected sectors should contact CISA or the FBI if compromise is found. MSPs should notify affected customers promptly.
What individuals should do
This story is mainly an enterprise problem, but it touches home users in one way. Compromised SOHO routers (TP-Link and Netgear were named as device classes) served as relay hops. Owners of such devices would rarely notice. Home and small-office users should install the latest firmware, change default admin credentials, disable remote administration, and replace routers that the manufacturer no longer supports. If a router is end-of-life, a factory reset does not fix the underlying risk.
FAQ
Is Silent Train a botnet?
Not in the traditional sense. Lumen describes short-lived chains of compromised devices assembled and torn down per operation. A Lumen researcher told CyberScoop the number of live nodes at any one moment is effectively close to zero. That is why there is no single network to take down.
Are all four zero-days patched?
Public reporting treats them as disclosed CVEs with vendor fixes, but this article cannot confirm exact fixed versions for each product. Check each vendor advisory. Also note that Lumen's "likely" Quest KACE zero-day was patched in October 2025, and its CVE is not identified in the reporting reviewed here.
Does this mean Volt Typhoon has already disrupted U.S. infrastructure?
No. Public reporting describes access and positioning, and government officials have characterized the activity as pre-positioning for a possible conflict. Lumen's findings do not show disruptive action.
Bottom line
This report matters because it shows a state actor investing in zero-days against the tools that administer other organizations' networks, while engineering its infrastructure to leave defenders almost nothing to find on the relay side. If you run an SD-WAN controller, RMM or similar platform, patch it, remove it from the internet, and hunt back through its logs. If you are an MSP or telecom, assume you are a priority target. Silent Train being "disrupted" does not mean the access it enabled has been found and removed.
Original reporting: The Hacker News
Protect yourself with tools recommended by cybersecurity professionals:
The tools below are independently selected based on security audits, transparency, and real-world effectiveness.