Cisco has released another wave of security updates for its Crosswork automation portfolio and Secure Workload Software, the product of what the vendor describes as an ongoing comprehensive internal security review. Nine vulnerabilities are addressed in this round. Five of them carry the maximum CVSS base score of 10.0. Four of the flaws affect Crosswork Data Gateway, Crosswork Network Controller, and Crosswork Planning — and critically, they are exploitable regardless of device configuration. There is no checkbox you can un-tick to be safe. Patching is the only path.
For those of us who have led IR engagements against compromised management planes, this advisory pattern should trigger immediate alarm. Crosswork is the automation backbone for large carrier and enterprise networks. Secure Workload (formerly Tetration) is the workload microsegmentation and telemetry platform. An unauthenticated attacker with CVSS 10.0 leverage against either platform is not just inside one box — they are inside the control plane that defines how your network routes, segments, and monitors everything else. That is a worst-case architectural position for a defender to lose.
Act this week. Inventory every Crosswork and Secure Workload instance, apply Cisco's fixes, and treat any internet- or broadly-reachable management interface as potentially compromised until proven otherwise.
Technical Analysis
Affected Products
Based on Cisco's disclosure, the nine vulnerabilities span:
- Cisco Crosswork Data Gateway — the collection and telemetry aggregation layer between network devices and Crosswork applications
- Cisco Crosswork Network Controller — the SDN orchestration and intent-based automation engine for WAN and transport networks
- Cisco Crosswork Planning — the network design, simulation, and capacity planning platform
- Cisco Secure Workload Software — the agent-based workload protection and microsegmentation platform deployed in on-premises form
Four of the vulnerabilities affect the Crosswork components independent of configuration state — meaning default and hardened deployments alike are exposed. Five of the nine flaws were scored CVSS 10.0, which in Cisco advisories almost universally denotes unauthenticated remote code execution, authentication bypass, or full privilege escalation reachable over the network.
Why CVSS 10.0 on a Management Plane Is Different
A perfect score on a user-facing application is bad. A perfect score on a network automation controller is catastrophic for a specific reason defenders must internalize:
- Crosswork holds privileged credentials for the network devices it manages (NETCONF/RESTCONF/SSH/telemetry subscriptions, SNMP, gNMI). Compromise of the controller is compromise of the managed fleet.
- Secure Workload agents run at high privilege on endpoints and servers to enforce segmentation. A compromised Secure Workload cluster can be used to weaken or disable enforcement policies fleet-wide — silently blinding the segmentation controls organizations depend on for lateral movement containment.
- Crosswork Data Gateway sits in-line with telemetry flows, making it an ideal persistence and traffic-manipulation point.
Exploitation Status
At the time of Cisco's disclosure, these flaws were identified through an internal security review — which is preferable to a post-incident disclosure, but should not breed complacency. Historically, the window between public disclosure of a CVSS 10.0 network-infrastructure flaw and in-the-wild weaponization is now measured in days, sometimes hours, as adversaries diff firmware and patches to reconstruct the vulnerable code paths. There is no confirmed public exploitation reporting in the source material yet, and no CISA KEV listing cited in the advisory summary — but that is a point-in-time statement, not a safety guarantee. Treat these as pre-weaponization criticals and patch on an emergency cadence.
Exploitation Requirements (Defender's View)
Without per-CVE technical detail in the disclosure summary, defenders should assume the most dangerous class: unauthenticated network-reachable exploitation of the web/API management plane. The practical exploitation preconditions to hunt against:
- Attacker has TCP reachability to the Crosswork or Secure Workload management interfaces (HTTPS/API ports)
- No valid credentials required for the five CVSS 10.0 issues
- Post-exploitation behavior likely manifests as: anomalous process execution on the appliance/container host, unexpected outbound connections from the management host, new local accounts or API tokens, and modification of managed-device credential stores
Detection & Response
Because Cisco has not published per-CVE indicators of compromise in the summarized disclosure, the correct defensive posture is behavioral detection on the management plane: web/API anomalies against the appliances, post-exploitation process execution on the Linux-based Crosswork/Secure Workload hosts, and egress anomalies from systems that should have highly predictable network behavior. Crosswork and Secure Workload appliances are among the most deterministic systems in your environment — they talk to known controllers, known managed devices, and Cisco infrastructure. Anything outside that baseline deserves scrutiny.
Sigma Rules
The following rules target post-exploitation behaviors on the Linux hosts underpinning Crosswork and Secure Workload deployments (these platforms run on Linux-based virtual/software form factors) and on egress anomalies. Tune the application path exclusions to your build versions — confirm legitimate service binary paths on a clean appliance before deployment.
---
title: Suspicious Shell Spawned from Cisco Crosswork or Secure Workload Service Processes
id: 3f8c1a2e-7b4d-4e91-a6c2-9d5f0e8b7a31
status: experimental
description: Detects interactive shells or command interpreters spawned by web, API, or Java service processes on Cisco Crosswork or Secure Workload hosts — a hallmark of post-exploitation activity following RCE against the management plane.
references:
- https://thehackernews.com/2026/08/cisco-patches-nine-crosswork-and-secure.html
- https://attack.mitre.org/techniques/T1059/004/
author: Security Arsenal
date: 2026/08/14
tags:
- attack.execution
- attack.t1059.004
- attack.initial_access
- attack.t1190
logsource:
category: process_creation
product: linux
detection:
selection_parent:
ParentImage|contains:
- 'java'
- 'nginx'
- 'apache'
- 'tomcat'
- 'node'
- 'crosswork'
- 'cndp'
- 'tetration'
- 'secureworkload'
selection_child:
Image|endswith:
- '/bash'
- '/sh'
- '/dash'
- '/zsh'
- '/python'
- '/python3'
- '/perl'
- '/curl'
- '/wget'
- '/nc'
- '/ncat'
- '/socat'
condition: selection_parent and selection_child
falsepositives:
- Legitimate automation scripts invoked by Cisco service daemons during upgrades or health checks — baseline during a known maintenance window
level: high
---
title: Credential or Configuration Store Access on Cisco Management Appliances
id: 91b2d4f6-3c8a-4e57-b2d9-1a6f7c0e5d42
status: experimental
description: Detects reads of credential, shadow, or application secret stores by non-system processes on Crosswork or Secure Workload hosts, consistent with post-compromise credential harvesting from the network controller.
references:
- https://thehackernews.com/2026/08/cisco-patches-nine-crosswork-and-secure.html
- https://attack.mitre.org/techniques/T1003/
- https://attack.mitre.org/techniques/T1552/
author: Security Arsenal
date: 2026/08/14
tags:
- attack.credential_access
- attack.t1003
- attack.t1552.001
logsource:
category: file_event
product: linux
detection:
selection:
TargetFilename|contains:
- '/etc/shadow'
- '/etc/passwd'
- '/etc/ssl/private'
- 'keystore'
- '.kube/config'
- 'secrets.yaml'
- 'credentials'
- 'application-secret'
filter_legit:
Image|endswith:
- '/sshd'
- '/sudo'
- '/login'
- '/systemd'
- '/kubelet'
condition: selection and not filter_legit
falsepositives:
- Backup agents and configuration management tools (Ansible, backup jobs) — restrict filter to approved automation service accounts
level: high
---
title: New Local Account Creation on Network Management Appliance
id: c47e9b1a-5d2f-4a83-8e6c-2b7d9f1a4e63
status: experimental
description: Detects user account creation via useradd/adduser or direct passwd modification on Cisco Crosswork or Secure Workload hosts — appliances where interactive account creation is rare and a strong persistence signal.
references:
- https://thehackernews.com/2026/08/cisco-patches-nine-crosswork-and-secure.html
- https://attack.mitre.org/techniques/T1136/001/
author: Security Arsenal
date: 2026/08/14
tags:
- attack.persistence
- attack.t1136.001
logsource:
category: process_creation
product: linux
detection:
selection:
Image|endswith:
- '/useradd'
- '/adduser'
- '/usermod'
- '/chpasswd'
filter_install:
CommandLine|contains: '--system'
condition: selection and not filter_install
falsepositives:
- Initial appliance provisioning and OEM service account creation during software upgrades — correlate against change tickets and patch windows
level: high
KQL — Microsoft Sentinel Hunting
Crosswork and Secure Workload management appliances can and should forward syslog/CEF to Sentinel. This query hunts for anomalous HTTP response patterns and outbound connections from management hosts — systems whose egress should be almost entirely to Cisco infrastructure, your NTP/DNS, and managed devices. Ingest appliance syslog via a Linux collector or your existing CEF pipeline.
// Hunt 1: Anomalous outbound connections from Crosswork / Secure Workload management hosts
let mgmt_hosts = dynamic ("crosswork", "cndg", "cwnc", "secureworkload", "tetration", "csw");
CommonSecurityLog
| where TimeGenerated > ago(7d)
| where DeviceProduct has_any ("Cisco") or SourceHostName has_any (mgmt_hosts)
| where isnotempty(DestinationIP)
| extend IsPrivateDest = ipv4_is_private(DestinationIP)
| where IsPrivateDest == false
| summarize ConnCount = count(), Destinations = make_set(DestinationHostName, 20), Ports = make_set(DestinationPort, 20)
by SourceHostName, DestinationIP, bin(TimeGenerated, 1h)
| order by ConnCount asc;
// Hunt 2: Suspicious HTTP request patterns against appliance management APIs (unauthenticated probing / exploitation attempts)
let mgmt_hosts2 = dynamic ("crosswork", "cndg", "cwnc", "secureworkload", "tetration", "csw");
Syslog
| where TimeGenerated > ago(7d)
| where HostName has_any (mgmt_hosts2)
| where SyslogMessage has_any ("/api/", "/crosswork/", "/app-ui/", "/oauth/", "/login", "/admin")
| where SyslogMessage has_any (" 401 ", " 403 ", " 500 ", " 404 ")
or SyslogMessage has_any ("../", "%2e%2e", "cmd=", "exec=", "bash", "wget", "curl", "${")
| summarize Attempts = count(), SampleRequests = make_set(SyslogMessage, 10)
by SourceIP = extract(@"(\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3})", 0, SyslogMessage), HostName, bin(TimeGenerated, 1h)
| order by Attempts desc;
// Hunt 3: Process execution anomalies forwarded from management appliance syslog
Syslog
| where TimeGenerated > ago(7d)
| where HostName has_any ("crosswork", "cndg", "cwnc", "secureworkload", "tetration", "csw")
| where SyslogMessage has_any ("useradd", "adduser", "/bin/bash", "/bin/sh", "chmod +x", "crontab", "systemctl enable", "authorized_keys")
| project TimeGenerated, HostName, SeverityLevel, SyslogMessage
| order by TimeGenerated desc;
Velociraptor VQL — Appliance Host Forensics
If you have Velociraptor (or equivalent agent-based collection) on the Linux hosts supporting Crosswork/Secure Workload VM deployments, this hunt sweeps for post-exploitation persistence and execution artifacts. Note: do not install third-party agents on Cisco appliance OS instances without confirming support implications — use this against your hypervisor-managed Linux VMs and collectors.
-- Hunt for post-exploitation artifacts on Cisco management-plane Linux hosts:
-- unexpected listeners, recently-modified persistence locations, and suspicious processes
-- Section 1: Processes with shells/interpreters spawned by service accounts
SELECT Pid, Ppid, Name, CommandLine, Exe, Username, CreateTime
FROM pslist()
WHERE (CommandLine =~ 'bash|/bin/sh|python|perl|nc |ncat|socat|wget|curl')
AND Username =~ 'root|nobody|www|tomcat|crosswork|tetration'
-- Section 2: Listening sockets that do not match expected management services
SELECT Pid, Name, Laddr, Lport, Status
FROM netstat()
WHERE Status =~ 'LISTEN'
AND NOT Lport IN (22, 80, 443, 8443, 161, 514, 6514, 30603)
-- Section 3: Recently modified persistence artifacts (last 7 days)
SELECT FullPath, Size, Mtime, Ctime
FROM glob(globs=[
'/etc/cron.d/*',
'/var/spool/cron/*',
'/home/*/.ssh/authorized_keys',
'/root/.ssh/authorized_keys',
'/etc/systemd/system/*.service',
'/tmp/*',
'/dev/shm/*'
])
WHERE Mtime > now() - (7 * 24 * 3600)
ORDER BY Mtime DESC
Remediation & Verification Script
Use the following on your Linux jump host / automation runner to inventory reachable Crosswork and Secure Workload management interfaces, confirm they are not exposed beyond approved management subnets, and verify post-patch service state. This is a verification-and-hardening script — patching itself is performed through Cisco's software update process per the advisory.
#!/bin/bash
# crosswork_secureworkload_verify.sh
# Verify exposure, versions, and hygiene of Cisco Crosswork / Secure Workload management planes
# Run from a management jump host with reachability to the appliance subnet(s).
set -euo pipefail
REPORT="cisco_mgmt_audit_$(date +%Y%m%d_%H%M%S).txt"
echo "=== Cisco Crosswork / Secure Workload Exposure & Hygiene Audit ===" | tee "$REPORT"
# 1) Define your appliance inventory (edit to match your environment)
APPLIANCES=("crosswork-nc.example.com" "crosswork-dgw.example.com" "crosswork-plan.example.com" "secureworkload.example.com")
MGMT_PORTS=(443 8443 30603 22)
# 2) Check management-plane reachability and TLS posture per appliance
echo -e "\n--- [1] Management Interface Reachability & TLS ---" | tee -a "$REPORT"
for host in "${APPLIANCES[@]}"; do
for port in "${MGMT_PORTS[@]}"; do
if timeout 5 bash -c "</dev/tcp/$host/$port" 2>/dev/null; then
echo "[OPEN] $host:$port" | tee -a "$REPORT"
if [ "$port" == "443" ] || [ "$port" == "8443" ]; then
echo | timeout 8 openssl s_client -connect "$host:$port" -servername "$host" 2>/dev/null \
| openssl x509 -noout -issuer -subject -dates 2>/dev/null | tee -a "$REPORT" || true
fi
else
echo "[CLOSED/FILTERED] $host:$port" | tee -a "$REPORT"
fi
done
done
# 3) Flag exposure: management interfaces should NEVER resolve/route from untrusted networks.
# From an external vantage (separate run), repeat section 2 — ANY open result = critical finding.
echo -e "\n--- [2] External Exposure Check ---" | tee -a "$REPORT"
echo "ACTION REQUIRED: Re-run this script from an untrusted network (or use an external scanner)." | tee -a "$REPORT"
echo "Any [OPEN] result on 443/8443/30603/22 from outside your mgmt VRF is a P1 finding." | tee -a "$REPORT"
# 4) Version verification — pull installed software version from the appliance CLI/API.
# Crosswork: SSH to the appliance and run 'show version' / check the UI System > Summary.
# Compare against the Fixed Software table in the Cisco Security Advisory.
echo -e "\n--- [3] Version Check (manual confirmation required) ---" | tee -a "$REPORT"
for host in "${APPLIANCES[@]}"; do
echo "-> SSH to $host and capture: show version (Crosswork) / System Info (Secure Workload UI)" | tee -a "$REPORT"
echo " Compare against Fixed Software releases: https://sec.cloudapps.cisco.com/security/center/publicationListing.x" | tee -a "$REPORT"
done
# 5) Audit local accounts and recent changes on supporting Linux infrastructure
echo -e "\n--- [4] Local Account & Persistence Audit (supporting Linux hosts) ---" | tee -a "$REPORT"
if [ -f /etc/passwd ]; then
echo "Accounts with interactive shells (UID >= 1000 or root):" | tee -a "$REPORT"
awk -F: '($3 >= 1000 || $3 == 0) && $7 ~ /(bash|sh|zsh)$/ {print $1" (uid "$3")"}' /etc/passwd | tee -a "$REPORT"
echo "Recently modified cron/systemd artifacts (7 days):" | tee -a "$REPORT"
find /etc/cron.d /var/spool/cron /etc/systemd/system -type f -mtime -7 2>/dev/null | tee -a "$REPORT" || true
fi
echo -e "\nAudit complete. Report: $REPORT"
Remediation
This is an emergency-change candidate, not a next-maintenance-window item. Five CVSS 10.0 flaws affecting configuration-independent deployment states means your compensating controls buy time — they do not close the hole.
- Inventory immediately. Identify every instance of Crosswork Data Gateway, Crosswork Network Controller, Crosswork Planning, and Secure Workload Software in your environment — including lab, staging, and DR instances, which are routinely forgotten and routinely softer targets.
- Patch now. Pull the fixed software releases directly from the Cisco Security Advisories portal. Match each advisory's "Fixed Software" table to your running version. Verify the upgrade completed and services restarted cleanly — a staged-but-not-applied patch leaves you exposed.
- Lock down the management plane as a compensating control. Crosswork and Secure Workload interfaces must be reachable only from a dedicated, ACL'd management VRF/subnet with jump-host enforcement. No internet exposure, ever. If any management interface is currently internet-reachable, treat that host as potentially compromised: isolate, capture forensic images/logs first, then rebuild and patch.
- Rotate credentials post-patch. Because these platforms store credentials for your managed network estate, rotate: device credentials held in Crosswork, API tokens and service accounts, Secure Workload API keys, and any local appliance accounts. If you suspect pre-patch exposure, this is not optional.
- Review Secure Workload enforcement policy integrity. Confirm segmentation policies have not been modified, no unauthorized policy scopes or enforcement-disabled workspaces exist, and agent health reporting matches your known fleet count. Silent policy tampering is the highest-value post-compromise move against this platform.
- Hunt retroactively. Run the Sigma/KQL/VQL content above against at least 30 days of retained telemetry where available. A CVSS 10.0 on a quiet management appliance is exactly the kind of flaw that sits exploited-but-unnoticed.
- Monitor for follow-on disclosure. Cisco noted this is part of a continued internal review — expect additional advisories. Subscribe to Cisco's security advisory feed and watch CISA KEV for additions related to these products.
Bottom Line
Management-plane compromise is the IR scenario that hurts the most and the longest, because it poisons the very systems you use to understand and control your environment. Nine flaws, five of them perfect scores, no configuration-based mitigation — this is as close to an unqualified "patch now" as advisories get. Get the fixes applied, get the management plane segmented, and assume the reconnaissance race between defenders and adversaries started the moment the advisory published.
Related Resources
Security Arsenal Penetration Testing Services AlertMonitor Platform Book a SOC Assessment vulnerability-management Intel Hub
Is your security operations ready?
Get a free SOC assessment or see how AlertMonitor cuts through alert noise with automated triage.