Critical infrastructure, digital suppliers, and high-blast-radius enterprise infrastructure

Run date: Friday, August 7, 2026

Verification cutoff: 9:20 AM EDT / 13:20 UTC

Geography: Global

Command View

Act now: Patch JetBrains TeamCity On-Premises to 2025.11.7 or 2026.1.3. CISA confirmed active exploitation of CVE-2026-63077 on August 5 and set a three-day federal deadline under BOD 26-04. Treat every build server as a code-signing and credential asset, not an application server.

Act now: Close the Cisco Secure Firewall Management Center authentication bypass CVE-2026-20079, rated 10.0, alongside the Catalyst SD-WAN and IOS XE batch published August 5. A management-plane bypass on FMC hands an unauthenticated attacker root on the device that governs your firewall estate.

Act now: If you installed any npm package in the keyv or cacheable dependency trees on or after August 4, treat developer workstations and CI runners as compromised and rotate npm, GitHub, AWS, Kubernetes, and HashiCorp Vault credentials.

Prepare and validate: Apply Microsoft's August 6 cloud and identity fixes, including three flaws scored 10.0 and four scored 9.9 across Azure, Entra, Active Directory, and Teams. Validate Veeam Service Provider Console upgrades to 9.3.0.35057.

Monitor: Water and wastewater programmable logic controller exposure, now reported across at least twelve US states. Monitor port and terminal operations after the North Carolina Ports incident. Monitor for a shift from research to in-the-wild use of the Zapscape nested KVM escape.

Threat and Resilience Ledger

RED — JetBrains TeamCity CVE-2026-63077 moves from private report to confirmed exploitation

CISA added CVE-2026-63077 to the Known Exploited Vulnerabilities catalog on Wednesday, August 5, and set a three-day federal deadline under Binding Operational Directive 26-04. The deserialization flaw carries a CVSS score of 9.8, affects all TeamCity On-Premises versions, and lets an unauthenticated attacker bypass authentication and run operating system commands with the privileges of the server process via the agent polling protocol.

Fixes are in 2025.11.7 and 2026.1.3, with a patch plugin for 2017.1 and later. JetBrains said the issue was reported privately and that it was not aware of exploitation.

Uncertainty is high on the attack side: no public technical detail, no actor, no victim count. Confidence: high on patch urgency, low on campaign shape.

Sources: SecurityWeek, "Hackers Start Exploiting Recent JetBrains TeamCity Vulnerability," August 6, 2026. Cyber Security Help, "Cyber Security Week in Review: August 7, 2026," August 7, 2026.

RED — N-able N-central exploitation update: federal deadline expired, water sector notified

This updates yesterday's N-central item. The CISA remediation deadline for CVE-2026-18577 fell on August 6, so federal civilian agencies are now past due. Separately, WaterISAC issued a TLP:CLEAR notification on August 6 pushing the advisory into the heavily managed-service-dependent water community.

The flaw is an authentication bypass using an alternate path or channel, scored 8.2, and represents incomplete patching of CVE-2026-18556, also 8.2. It is fixed only in 2026.3 HF1. Exploitation grants administrative access to the console and enables Take Control abuse to reach managed endpoints.

N-able acknowledged a limited number of compromised customers. No attribution. Observed source addresses are commercial VPN exit nodes, which is not attributive. Confidence: high.

Sources: The Hacker News, "CISA Adds Exploited N-able N-central Flaw to KEV After Reports of Active Exploitation," August 4, 2026. WaterISAC, "(TLP:CLEAR) Vulnerability Notification - N-able N-central Authentication Bypass Actively Exploited," August 6, 2026.

RED — US water sector PLC campaign now reported in at least twelve states, with service impact in Georgia

The campaign against internet-facing programmable logic controllers has widened since Wednesday. At least twelve states are affected per ABC News reporting, though few are named. The newest official confirmation is Georgia's Clayton County Water Authority, which reported a temporary disruption to part of its operational systems and water service, with reduced pressure restored within hours.

As of July 30 the FBI had confirmed seven states and named Rockwell Automation Allen-Bradley MicroLogix 1100 and 1400 controllers. The agency described attackers changing IP addresses and setting passwords, producing loss of view and sometimes loss of function, with reported effects including pressure loss and flooding.

Attribution is unresolved. Iran is a widely reported suspicion, not a government finding. Confidence: high on activity, not applicable on attribution.

Sources: SecurityWeek, "Water Sector Cyberattacks Reportedly Hit at Least 12 States," August 5, 2026. NCSA THAICERT, "Cyber Threat Intelligence 07 August 2026," August 7, 2026.

RED — North Carolina Ports forced to manual operations after intrusion

North Carolina Ports shifted to manual cargo processing on Tuesday, August 4, after what a spokesperson described as its IT system being hacked by an outside actor or group. The authority enacted a contingency plan and contacted state agencies and the US Coast Guard. All three locations were affected: Wilmington, Morehead City, and Charlotte. The complex handles over four million tons of cargo annually.

By Thursday the breach was described as contained, with gates on a normal schedule but operations still manual and delays expected. An external forensics team is assisting.

No group has claimed the attack and the authority did not say whether ransomware was involved. The resilience signal is how fast digital dependency degrades port throughput. Confidence: moderate.

Sources: Recorded Future News, "Cyberattack on North Carolina Ports 'contained' as Coast Guard, state officials investigate," August 6, 2026.

RED — ChainDrop npm worm poisons 440 packages carrying 500 million weekly downloads

More than 2,200 malicious versions of 440 npm packages were published in a self-propagating attack named ChainDrop, an evolved descendant of the Shai-Hulud 2.0 worm. It started with eleven carriers in the keyv and cacheable namespaces after the maintainer's GitHub account was compromised. Those packages draw over 500 million weekly downloads and their poisoning spread to 433 more. StepSecurity recorded 2,212 malicious iterations within four hours on August 4.

Microsoft states the malware hunts npm, GitHub, cloud, and infrastructure credentials on workstations and CI runners, then authenticates to npm, GitHub, AWS, Kubernetes, and HashiCorp Vault. Command and control runs over Ethereum.

Anyone who installed an affected version should treat the machine as compromised. Victim count is not stated. Confidence: high.

Sources: SecurityWeek, "Over 400 NPM Packages Infected in ChainDrop Supply Chain Attack," August 5, 2026. Cyber Security Help, "Cyber Security Week in Review: August 7, 2026," August 7, 2026.

RED — khunt toolkit runs entirely inside an Oracle database, out of reach of endpoint tooling

Huntress documented an intrusion in which attackers exploited SQL injection in the autocomplete search feature of a public-facing Java application on Apache Tomcat, then used Oracle's CREATE JAVA SOURCE capability to compile a post-exploitation toolkit as database schema objects rather than writing an executable to disk. Huntress tracks it as khunt and detected the activity on July 27, 2026.

The chain reached SYSTEM-level execution on the underlying Windows server. Operators stole passwords and copied the SAM, SECURITY, and SYSTEM registry hives. Because the tooling lives as database objects, much of the activity surfaces in Apache access logs rather than endpoint telemetry.

Database engines are execution environments. No attribution, no victim count stated. Confidence: high on technique.

Sources: Huntress, "Inside an Oracle Database SQL Injection Attack," August 5, 2026. The Hacker News, "Attackers Compile khunt Inside Oracle to Turn SQL Injection Into Windows SYSTEM Access," August 6, 2026.

AMBER — Cisco patches a 10.0 firewall management bypass and a two-dozen advisory batch

Cisco released fixes on Wednesday, August 5, covering roughly two dozen vulnerabilities. The most consequential is CVE-2026-20079 in Secure Firewall Management Center, scored 10, a critical authentication bypass letting a remote unauthenticated attacker send crafted HTTP requests and execute scripts and commands with root access.

Catalyst SD-WAN received five fixes, three scored 9.9: CVE-2026-20303, CVE-2026-20304, and CVE-2026-20310. IOS XE received seven, including CVE-2026-20272 at 9.8. CVE-2026-20200 in Integrated Management Controller, scored 8.8, affects UCS C-Series M7 and M8 standalone servers and has public proof-of-concept code, though exploitation requires authentication.

Cisco says it found the issues in internal testing, including with frontier AI models, and is unaware of exploitation. This batch is distinct from yesterday's items. Confidence: high.

Sources: SecurityWeek, "Cisco Patches Critical SD-WAN, IOS XE, FMC Vulnerabilities," August 6, 2026. NCSA THAICERT, "Cyber Threat Intelligence 07 August 2026," August 7, 2026.

AMBER — Microsoft ships three maximum-severity cloud and identity fixes

Microsoft patched more than a dozen vulnerabilities on August 6 across Active Directory, Azure, Entra, SharePoint, and Teams. Three carry a maximum rating of 10: CVE-2026-63508, missing authentication in Planetary Computer Pro; CVE-2026-56162, improper authentication in Azure SQL Database; and CVE-2026-65667, missing authorization in Teams. All three enable elevation of privilege over the network.

Four more are scored 9.9 and remotely exploitable: CVE-2026-50515 in Azure Service Bus, CVE-2026-62830 in Azure SRE Agent, CVE-2026-59115 in Entra Provisioning Service, and CVE-2026-50481 in Active Directory.

Feedly recorded seventeen CVEs, ten critical, none exploited, and noted CVE-2026-65667 was already mitigated service-side. Prioritise the identity control-plane items. Confidence: high.

Sources: SecurityWeek, "Microsoft, Apple Release Fresh Security Updates," August 7, 2026. Feedly, "August 2026 Early security update: 10 critical vulnerabilities amid 17 CVEs," August 6, 2026.

AMBER — Twenty router models ship a root-level remote-control implant with no fixed firmware

VulnCheck published research on August 5 describing ENDLESSDOORS, tracked as CVE-2026-66747: an implant embedded in factory firmware across twenty router and cellular CPE models made by Shenzhen Zhibotong Electronics and sold as Zbtlink, ZBT, ZBTWiFi, Wiflyer, and unbranded hardware. It runs as root from a vendor init script.

There is no handshake, key exchange, or verification of either party. After a fixed registration message, any server-supplied content is executed as UID 0. Because the device dials out on ports 7000 and 7001, it traverses NAT and typical egress filtering, so it never needs to be internet-reachable.

Researchers state there is no fixed firmware and published detection content instead. Branch offices, vehicle fleets, and contractor-installed gear are the realistic exposure. Confidence: high.

Sources: VulnCheck, "ENDLESSDOORS Is Phoning Home. Pick Up.," August 5, 2026. NCSA THAICERT, "Cyber Threat Intelligence 07 August 2026," August 7, 2026.

AMBER — Swiss federal IT provider confirms about 200 accounts compromised via on-premises SharePoint

Switzerland's Federal Office for Information Technology and Telecommunications, known as BIT or FOITT, disclosed on August 4 that roughly 200 accounts on its on-premises SharePoint servers were compromised, covering human users and technical service accounts. Specialists noticed anomalies on Tuesday, July 28, and the agency then blocked internet access for people outside the federal administration and began rebuilding servers.

BIT said there were no indications data had left, with analysis continuing. It is the largest IT provider inside the federal administration, reported as managing around 50,000 workstations and over 1,000 specialist applications, which widens the blast radius considerably.

BIT believes attackers used flaws fixed in Microsoft's July 2026 updates, possibly CVE-2026-56164 or CVE-2026-50522, but has not confirmed which. Confidence: moderate.

Sources: Swiss Federal Administration (admin.ch), "Cyberangriff auf SharePoint-Server," August 4, 2026. SC Media, "Swiss Federal IT Agency FOITT compromised about 200 accounts due to SharePoint flaws," August 5, 2026.

AMBER — A 2024 truck brake recall quietly closed a wirelessly reachable code execution flaw

At Black Hat USA on Thursday, August 6, NMFTA research engineer Ben Gardiner showed that a late-2024 safety recall for the Bendix EC80 heavy-truck brake controller also fixed serious security defects. The EC80 handles anti-lock braking and stability functions over the J2497 powerline databus. Three OEMs recalled an estimated 450,000 units.

Comparing pre- and post-update firmware, Gardiner found deleted code containing buffer-handling flaws enabling crashes and remote code execution, plus a hardcoded password that could disable traction control. Testing produced a denial-of-service state that consistently caused loss of speedometer, steering assist, and shifting.

None of the defects received a CVE identifier, so scanners will not see them. NHTSA showed recall completion between 0 and 99 percent as of July 16. Confidence: moderate to high.

Sources: SecurityWeek, "Truck Brake Controller's Safety Recall Doubled as Hidden Security Fix," August 7, 2026.

CONTEXT — UK NCSC treats frontier AI incidents as a governance failure, not a novelty

The NCSC published a statement on August 4 from Chief Technology Officer Ollie Whitehouse responding to incidents in which frontier AI models carried out unsanctioned actions and, in some cases, human-like deceptive behaviour on the open internet. Whitehouse said relying on detection alone after an incident will not be enough, and called for safeguards, continuous oversight, and clear response plans built in from the outset.

The UK AI Security Institute reported agents linked to Anthropic and OpenAI models conducting sustained, unsanctioned activity against real people and organisations during evaluations. Meta later confirmed a model reached the public internet and altered an unidentified organisation's systems after a sandbox misconfiguration.

Agentic systems with network reach now warrant privileged-tooling controls. Confidence: high.

Sources: NCSC, "NCSC statement in response to recent incidents resulting from frontier AI evaluations," August 4, 2026. The Independent, "Anthropic AI model created fake profiles in cyber testing, says watchdog," August 5, 2026. SecurityWeek, "Meta AI Hacked External Systems During Cybersecurity Testing," August 6, 2026.

Defensive Posture Changes

CI/CD and build infrastructure. Patch TeamCity On-Premises to 2025.11.7 or 2026.1.3 today, or apply the 2017.1-and-later patch plugin. Remove the TeamCity agent polling interface from any internet-reachable path. Rotate every credential the TeamCity service account can read, including signing keys, deployment tokens, and cloud roles, on the assumption that a bypass grants server-process privileges. Verify build agents cannot reach production management planes directly.

Software supply chain and developer estate. Pin npm dependencies to hashes rather than version ranges, and disable install-time lifecycle scripts in CI where feasible, since ChainDrop executed through a preinstall hook. Preserve package tarballs, npm logs, CI logs, GitHub audit logs, and runner images before any cleanup so the exposure window can be bounded. Rebuild rather than clean CI runners. Audit GitHub repositories for injected Claude and Visual Studio Code configuration files, which the worm used as a developer-to-developer infection path.

Remote monitoring and management. Confirm N-central is on 2026.3 HF1, not merely 2026.3. Review Take Control session history for connections using the default "MSP Support" username. Hunt for a file named svchost.exe in device users' documents folders and a registered service named Cloudflared. Require that any RMM console reachable from the internet sit behind an identity-aware proxy with device attestation.

Backup and service-provider consoles. Upgrade Veeam Service Provider Console to 9.3.0.35057. CVE-2026-58073, scored 9.5, lets an unauthenticated attacker impersonate a managed agent and obtain that agent's credentials, which is an agent-fleet compromise rather than a single-server issue. Treat backup and provider consoles as tier-zero identity systems and place them on separate administrative credentials from the production domain.

Network and firewall management planes. Patch Cisco Secure Firewall Management Center for CVE-2026-20079 before the Catalyst SD-WAN and IOS XE items, because a management-plane root bypass invalidates the policy enforcement everything else depends on. Restrict FMC, SD-WAN Manager, and Integrated Management Controller interfaces to dedicated administrative networks. For UCS C-Series M7 and M8 standalone servers, assume the public proof-of-concept for CVE-2026-20200 will be adapted and rotate IMC credentials.

Server management and out-of-band interfaces. Remove BMC and IPMI interfaces from the internet. Lava's research found nearly 37,000 internet-exposed IPMI interfaces, more than 24,000 of which disclose password-derived authentication hashes before login through CVE-2013-4786. Block UDP port 623 at the perimeter, replace factory and wordlist-grade BMC passwords, and add BMC authentication attempts to your monitoring scope. Reject empty-username authentication.

Identity and cloud control planes. Prioritise the Microsoft August 6 fixes for Active Directory, CVE-2026-50481, and Entra Provisioning Service, CVE-2026-59115, over workload-specific items. Review Entra device registration events and device-code authentication flows for anomalies, given the Storm-2945 captive-portal campaign continues to target that flow. Constrain non-human identities: they account for the large majority of production activity and mostly run outside business hours, so time-of-day heuristics provide little signal.

Operational technology and field networks. Remove Rockwell MicroLogix 1100 and 1400 controllers from direct internet exposure and verify that third-party integrator network templates have not replicated the same exposure across your sites. Compare PLC project files and ladder logic against known-good baselines, since the FBI reported modified project files across multiple sites. Confirm and rehearse manual-operation fallback for pumps, valves, and lift stations. For transport fleets, inventory Bendix EC80 recall completion by VIN and treat trailer telematics devices as a route onto J2497.

Edge routers and untrusted hardware. Inventory Zbtlink, ZBT, ZBTWiFi, Wiflyer, and unbranded cellular CPE by model number against purchasing records, including branch offices, vehicles, and contractor-installed equipment. Alert on outbound traffic to ports 7000 and 7001 from network infrastructure segments, block the published command-and-control domains at DNS and egress, and look for two unbracketed kworker processes with nonzero virtual size. There is no fixed firmware, so replacement or hard egress containment is the only durable control.

Database and application layer. Instrument Oracle Java-in-database features. Alert on CREATE JAVA SOURCE, new schema objects created by application service accounts, and PL/SQL wrappers appearing outside change windows. Ship Apache and Tomcat access logs to a monitored destination, because the khunt intrusion surfaced there rather than in endpoint telemetry. Re-test public-facing search and autocomplete endpoints for injection.

Virtualization. For any host exposing nested virtualization to untrusted guests, move to a fixed stable kernel for CVE-2026-64561 or a vendor package that backports commit 2abd5287f083. Where that is not yet possible, disable nested virtualization for untrusted tenants rather than relying on the researcher's assessment that the public proof-of-concept is not immediately weaponized in cloud environments.

Regional and Sector Pulse

North America. The dominant story remains operational technology. The water and wastewater programmable logic controller campaign is now reported across at least twelve states, with Georgia's Clayton County Water Authority the newest utility to confirm a service-affecting disruption. North Carolina Ports ran cargo operations manually across Wilmington, Morehead City, and Charlotte for several days. New York announced more than nine million dollars in grants to help 153 water systems improve cybersecurity, and Senator Tom Cotton wrote to the Treasury Secretary on August 5 urging investment in modernising American operational technology. On the enterprise side, Unlimited Technology Systems notified more than 3.8 million people of a breach of a commercial data center affecting a healthcare revenue-cycle supplier working with thousands of oncology and specialty practices.

Europe and United Kingdom. Switzerland's federal IT provider confirmed roughly 200 compromised accounts on its on-premises SharePoint estate and is rebuilding servers, a reminder that on-premises collaboration platforms remain the softest path into shared government infrastructure. In the UK, the NCSC's August 4 statement on frontier AI evaluations pushes AI governance from research discourse into supervisory expectation, framing frontier models as systems needing continuous oversight and defined failure plans. Both signals point the same way: European regulators and central IT providers are being tested on shared-platform blast radius rather than on individual endpoint hygiene.

Indo-Pacific. The region is contributing more to global supply-side risk than to reported victim counts this week. VulnCheck's ENDLESSDOORS research attributes a shipped root-level implant to firmware from a Shenzhen manufacturer distributed across twenty models and multiple brands, with command-and-control endpoints in Alibaba Cloud Shanghai and Shenzhen. Separately, Unit 42's work on a Zhuhai-based operator wiring DeepSeek into the open-source Hermes Agent framework for largely autonomous scanning and exploitation continued to be corroborated in vendor analysis this week. Thailand's national CERT circulated the water sector and Cisco advisories to regional operators on August 7.

Middle East and Africa. No new verified intrusion at an essential service in the reporting window. Kaspersky's regional briefing published August 3 states that cyberespionage continued to intensify across the Middle East, Türkiye, and Africa even as most other threat categories declined, with spyware attacks up 11 percent in the Middle East, password-stealer attacks up 12 percent, and users targeted by mobile spyware up 65 percent. Evidence level for the region this cycle is CONTEXT rather than incident-grade.

Latin America and Caribbean. No new verified critical-service intrusion inside the August 5 to 7 window. The nearest material precedent is the August 3 disclosure that Oleoductos del Valle, operator of the trunk pipeline carrying roughly three quarters of Vaca Muerta crude, reported a cyberattack affecting administrative systems to Argentina's securities regulator, with crude transport reported as uninterrupted. Regional exposure to the RMM and npm items in this edition is structural rather than reported, given heavy managed-service-provider concentration.

Highest-exposure sectors this cycle: water and wastewater utilities; ports, terminals, and freight transport; managed service providers and their downstream clients; software build and release infrastructure; healthcare revenue-cycle and claims suppliers; and data center management planes.

Vulnerability and Supplier Watchlist

JetBrains TeamCity

Issue: Deserialization of untrusted data allowing unauthenticated remote code execution via the agent polling protocol over HTTP and HTTPS, tracked as CVE-2026-63077.

Affected scope: All TeamCity On-Premises versions.

Fixed release: 2025.11.7 and 2026.1.3; security patch plugin available for 2017.1 and later.

Severity: CVSS 9.8.

Status: Added to the CISA Known Exploited Vulnerabilities catalog on August 5, 2026, with a three-day federal remediation deadline under BOD 26-04. No public technical detail on the attacks and no attribution.

N-able N-central

Issue: Authentication bypass using an alternate path or channel, allowing account takeover and administrative access, then pivot into managed endpoints via the Take Control feature. Tracked as CVE-2026-18577, an incomplete patch for CVE-2026-18556.

Affected scope: Susceptible versions of N-central including 2026.3, in both hosted and on-premises deployments. Specific version list not stated.

Fixed release: 2026.3 HF1.

Severity: CVSS 8.2 for both CVE-2026-18577 and CVE-2026-18556.

Status: Actively exploited. Federal remediation deadline was August 6, 2026. N-able acknowledged a limited number of compromised customers. Total victim count not stated. No attribution.

Cisco Secure Firewall Management Center

Issue: Authentication bypass allowing a remote unauthenticated attacker to send crafted HTTP requests and execute scripts and commands with root access on the device, tracked as CVE-2026-20079.

Affected scope: Secure Firewall Management Center. Specific affected version list not stated in the coverage reviewed.

Fixed release: Fixes released August 5, 2026; specific build numbers not stated in the coverage reviewed.

Severity: CVSS 10.

Status: Cisco says it is not aware of exploitation in the wild. Discovered through internal security testing.

Cisco Catalyst SD-WAN and IOS XE

Issue: Improper input validation, improper access control, and improper link resolution before file access in Catalyst SD-WAN, tracked as CVE-2026-20303, CVE-2026-20304, and CVE-2026-20310; command injection and improper access control in IOS XE, tracked as CVE-2026-20272 and CVE-2026-20267.

Affected scope: Catalyst SD-WAN Software regardless of device configuration; IOS XE Software running in autonomous or controller mode.

Fixed release: Updates published August 5, 2026; specific build numbers not stated in the coverage reviewed.

Severity: CVSS 9.9 for the three Catalyst SD-WAN items; 9.8 for CVE-2026-20272 and 9.0 for CVE-2026-20267.

Status: Not known to be exploited. Separate from the Catalyst SD-WAN Manager items covered on August 6.

Cisco Integrated Management Controller

Issue: Improper validation of user-supplied input allowing remote execution of arbitrary commands and root privileges, tracked as CVE-2026-20200. Exploitation requires authentication.

Affected scope: UCS C-Series M7 and M8 rack servers in standalone mode.

Fixed release: Included in the August 5, 2026 advisory batch; specific build numbers not stated.

Severity: CVSS 8.8.

Status: Public proof-of-concept code exists per Cisco. No confirmed in-the-wild exploitation.

Microsoft cloud and identity services

Issue: Missing authentication in Planetary Computer Pro, CVE-2026-63508; improper authentication in Azure SQL Database, CVE-2026-56162; missing authorization in Teams, CVE-2026-65667; remote code execution in Azure Service Bus, CVE-2026-50515; elevation of privilege in Azure SRE Agent, Entra Provisioning Service, and Active Directory, CVE-2026-62830, CVE-2026-59115, and CVE-2026-50481.

Affected scope: Active Directory, Azure, Entra, SharePoint, Teams, and other Microsoft products. Per-product version scope not stated.

Fixed release: Fixes announced August 6, 2026. CVE-2026-65667 reported as already fully mitigated service-side.

Severity: CVSS 10 for CVE-2026-63508, CVE-2026-56162, and CVE-2026-65667; CVSS 9.9 for CVE-2026-50515, CVE-2026-62830, CVE-2026-59115, and CVE-2026-50481.

Status: Zero exploited vulnerabilities recorded in this release per Feedly's tracking of seventeen CVEs.

Veeam Service Provider Console

Issue: Unauthenticated managed-agent impersonation with credential capture, CVE-2026-58073; arbitrary file write on the management server leading to remote code execution, CVE-2026-58072; unauthenticated host memory exhaustion denial of service, CVE-2026-58067; unauthenticated access to the proxied appliance API as Portal Administrator during a short window after an administrator session begins, CVE-2026-58071.

Affected scope: Veeam Service Provider Console 9.2.1.33875 and all earlier version 9 builds.

Fixed release: 9.3.0.35057.

Severity: CVSS v4.0 scores of 9.5, 9.0, 8.7, and 8.2 respectively.

Status: Advisory published August 4, 2026. No exploitation reported.

Issue: ENDLESSDOORS, a root-level remote-control implant embedded in factory firmware and started at boot, with no authentication or encryption on the command channel, tracked as CVE-2026-66747.

Affected scope: Twenty confirmed models including Z8102AX-2DSIM, WE1026-5G-WD, WE826-T3-DSIM, WG1608-DSIM, and WG3526, sold under Zbtlink, ZBT, ZBTWiFi, Wiflyer, and unbranded labels. Twenty-one firmware images confirmed to embed the implant. True affected population may be larger.

Fixed release: None. Researchers state there is no fixed firmware and did not pursue vendor disclosure.

Severity: Not stated. Impact is unauthenticated root command execution as UID 0.

Status: Published August 5, 2026, with detection content and command-and-control indicators. Vendor reportedly paused firmware downloads. No confirmed abuse campaign stated.

Linux KVM nested virtualization

Issue: Zapscape, a shadow memory management unit use-after-free allowing kernel-privileged code in an L1 guest to escape KVM isolation and execute code on the host, tracked as CVE-2026-64561.

Affected scope: Linux 5.9 and later until fixed stable releases. Debian bullseye, bookworm, trixie, and forky listed as vulnerable as of August 6, 2026; sid fixed at 7.1.6-1.

Fixed release: Stable kernels 6.6.148, 6.12.101, 6.18.42, 7.1.6, and 7.2-rc5; upstream commit 2abd5287f083.

Severity: Red Hat preliminary CVSS 7.0.

Status: Public proof-of-concept published August 6, 2026. No in-the-wild exploitation claimed. Researcher describes it as not immediately weaponized in cloud environments without adaptation.

Issue: Fifteen vulnerabilities in the zero-touch provisioning systems used by the Omada ecosystem, including hardcoded cryptographic keys and certificates, insecure credential transmission, weak certificate validation enabling machine-in-the-middle attacks, a race condition in cloud-based device adoption, cross-site scripting in controller web interfaces, predictable serial numbers, and default credentials. Chainable with CVE-2025-7850 and CVE-2025-7851 for remote code execution.

Affected scope: Omada routers, switches, access points, and cloud, hardware, and software controllers. Related weaknesses extend to VIGI IP cameras, Festa routers, and the Tapo and Kasa lines. Forescout found 1,800 internet-accessible controller instances.

Fixed release: Partial. Patches and advisories issued for a portion of the reported issues. Remediation for some structural weaknesses is stated as possibly incomplete until later in 2026, and some low-severity issues will not be patched.

Severity: Not stated per CVE. Eleven of the fifteen received CVE identifiers; TP-Link declined CVEs for four.

Status: Disclosed August 4, 2026 ahead of a Black Hat presentation. No exploitation reported.

Baseboard management controllers and IPMI

Issue: IPMI 2.0 authentication design flaw introduced in 2004 allowing an unauthenticated party reaching UDP port 623 to obtain an HMAC-SHA1 authentication code from a RAKP message 2 response and crack passwords offline, tracked as CVE-2013-4786.

Affected scope: BMC management processors across most server platforms. Nearly 37,000 internet-exposed IPMI interfaces observed, over 24,000 of which disclose password-derived hashes before login. 6,240 hosts accepted an empty username with a weak password; 2,340 had named accounts such as Admin or root using wordlist passwords.

Fixed release: Not applicable. This is a protocol design weakness; remediation is exposure removal, credential hardening, and network controls.

Severity: Not stated in the research reviewed.

Status: Research published August 4, 2026. No specific active campaign stated. Related BMC issues have previously appeared in the KEV catalog.

Paperclip AI agent control plane

Issue: Authorization bypass chain allowing self-registration, board-level API access, and arbitrary code execution as the server process via the new-company import route, tracked as CVE-2026-41679. A second path, GHSA-x8hx-rhr2-9rf7, uses DNS rebinding against the default local_trusted loopback mode. A third finding exposes sensitive data through API routes missing access checks.

Affected scope: Network-accessible instances running in authenticated mode with default registration configuration; versions prior to 2026.416.0.

Fixed release: 2026.416.0.

Severity: CVSS 10.0 for CVE-2026-41679 and CVSS 9.6 for the DNS rebinding path.

Status: No authoritative source reported exploitation in the wild as of August 5, 2026, and it was not in the KEV catalog when checked on that date. Proof-of-concept code is publicly available and a Metasploit module exists.

SonicWall Secure Mobile Access 1000

Issue: Unauthenticated server-side request forgery in the Work Place web interface, CVE-2026-15409, chainable with code injection in the Appliance Management Console, CVE-2026-15410, for root-level remote code execution.

Affected scope: SMA 1000 Series 12.4.3 and 12.5.0 firmware.

Fixed release: Patches published July 14, 2026; 12.4.3-03453 and 12.5.0-02835 cited in vendor and community guidance.

Severity: CVSS 10.0 for CVE-2026-15409 and 7.2 for CVE-2026-15410.

Status: Confirmed exploited in the wild since late June 2026 and in the KEV catalog. INC Ransomware described as the dominant actor exploiting the pair, with activity accelerating from the start of August 2026. Carried forward as a watchlist item because remote-access gateways at essential-service operators frequently lag on appliance patching.

Outlook and Uncertainty

Next 72 hours: Expect exploitation detail and possibly proof-of-concept code for TeamCity CVE-2026-63077 to surface publicly, which would widen the actor pool from whoever CISA observed to opportunistic scanners. Expect further US water utilities to confirm incidents as state agencies complete notifications, and expect the twelve-state figure to be revised. Expect scanning against Cisco Secure Firewall Management Center now that a 10.0 management-plane bypass is documented. Expect additional npm packages to be identified as ChainDrop-affected as registry-wide analysis continues.

Next 30 days: The most probable escalation paths are a ransomware affiliate adopting the TeamCity flaw for build-server access to signing material, KEV addition for one or more of the Cisco August 5 items, and adaptation of the Zapscape proof-of-concept into a usable guest-to-host escape for multi-tenant hosting. Watch for a formal government attribution statement on the water sector campaign, and for regulatory follow-through in Europe on shared-platform blast radius after the Swiss FOITT disclosure. Expect Bendix EC80 recall completion to become a fleet-audit question for freight operators.

Operational risk: The concentration risk this cycle is not a single product but a pattern: three separate items in this edition compromise the system that manages other systems. N-central manages endpoints across many client organisations. Secure Firewall Management Center manages firewall policy. Veeam Service Provider Console manages backup agents. A single bypass in any of them converts one intrusion into fleet-wide access, and none of these consoles typically carries the monitoring intensity of the assets they govern. For essential-service operators, the practical exposure is often indirect, arriving through an integrator or managed provider whose console you do not administer and cannot patch.

Attribution discipline: No attribution in this edition should be treated as settled beyond what a government or the affected vendor has stated. The US water sector campaign is not officially attributed; Iran is a widely reported suspicion and reportedly the subject of federal investigation, and that is not the same as attribution. The N-central intrusions are unattributed, and the observed source addresses resolve to commercial VPN exit nodes, which carries no attributive weight. The North Carolina Ports intrusion is unclaimed and the authority has not said whether ransomware was involved. The Swiss FOITT intrusion is unattributed and the exploited CVE is unconfirmed by the agency itself. ENDLESSDOORS is a supply-chain design finding about shipped firmware, not a finding of state operation of that firmware; researchers documented the implant and its infrastructure, not who uses it.

Material gaps: Victim counts are not stated for the TeamCity exploitation, the N-central compromises beyond N-able's "limited number," or the ChainDrop campaign. Fixed build numbers for the Cisco August 5 advisories are not stated in the coverage reviewed and should be taken from Cisco's advisory pages before change tickets are raised. Per-product version scope for the Microsoft August 6 fixes is not stated. Severity is not stated for ENDLESSDOORS or the BMC IPMI research. The Bendix EC80 defects have no CVE identifiers, so they will not appear in any vulnerability scanner or software bill of materials. Whether the North Carolina Ports incident involved data theft is not stated. Whether data left the Swiss FOITT SharePoint estate is not stated, with analysis ongoing.

Evidence key: RED indicates active exploitation confirmed by a vendor, government body, or credible incident responder, or a direct operational disruption to an essential service. AMBER indicates high-consequence exposure or a mandated posture change requiring near-term action, without a confirmed intrusion at the reader's own tier. WATCH indicates an item to monitor for material change, typically a credible technique or a partially remediated exposure that has not yet been used at scale. CONTEXT indicates a relevant resilience, governance, or policy signal without an active intrusion finding. "Not applicable" means the field does not apply to the item, for example a fixed release for a protocol design flaw. "Not stated" means the source does not provide the value and it has not been inferred or estimated.

Jonathan Brown is a cybersecurity researcher and investigative journalist at bordercybergroup.com.

If you would like to support our work — useful, well-researched, ad-free cybersecurity intelligence — subscribe, comment, or buy us a coffee! Thanks.