SecurityNews

Patchwork – Wednesday 18 November

This edition leads with Java4Shell, a critical flaw in Java’s Base64 decoder that lets attackers execute code without logging in on Jira, Confluence, Jenkins and Keycloak, among others. In addition, the websites of KLM, ABN AMRO, ICS Cards and more than 170 other organisations served malicious code for a day via the marketing platform Optimizely, N-able releases Hotfix 3 for N-central with seventeen new security fixes, a flaw in Puppet Enterprise is being actively exploited, and Zyxel patches a critical flaw in modems and Wi-Fi extenders that are also used by Dutch providers.

Java4Shell: critical Java flaw gives attackers unauthenticated access to Jira, Confluence, Jenkins and Keycloak

Vulnerability Memory error (out-of-bounds write) in the AVX-512-accelerated Base64 decoder of the HotSpot JVM. Affected are Java 17.0.6 and later, Java 21, 25 and 27 of all OpenJDK distributions, Oracle Java SE 8u511-perf and Oracle GraalVM, only on x86-64 processors with AVX-512 VBMI (including Intel Ice Lake and newer, AMD Zen 4 and newer). Standard Java 8 and 11 and ARM64 servers are not affected
CVE number CVE-2026-114938
CVSS score 8.1 according to Oracle (CVSS 3.1); 9.2 according to discoverer Calif (CVSS 4.0)
Actively exploited No, no exploitation known as of 18 November. The technique is public, but Calif is withholding the product-specific exploits for now
Patch available Yes, partially. Oracle Java SE 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1 and 8u513-perf; Amazon Corretto and Azul Zulu also have fixes. Red Hat, Eclipse Temurin and Microsoft Build of OpenJDK did not yet have a fixed build at the time of publication. Temporary workaround: the JVM option -XX:-UseBASE64Intrinsics

How can the vulnerability be exploited?

Oracle fixed the flaw on 17 November in its November Critical Security Patch Update and credits Quang Luong of security firm Calif with the discovery. According to Qualys, when given a short, malformed Base64 message the decoder writes 62 bytes past the end of a buffer, before Java even rejects the input. Because Base64 is used everywhere in logins and tokens, this can happen before authentication. According to The Hacker News, Calif has demonstrated unauthenticated code execution in the default configurations of Keycloak, Jira, Confluence, JFrog Artifactory, Jenkins and PingFederate, and post-authentication in Kafka, Solr and Elasticsearch, among others. Calif says a Java program connecting to a malicious TLS server can also be taken over.

What could the impact be?

An attacker can read secrets such as keys and tokens from memory, execute code or crash the JVM. There is no universal exploit: each product requires its own attack chain, which slows down mass exploitation. The tricky part for administrators is the inventory. CISA points out that Java is often bundled in third-party software, container images and appliances, and therefore does not appear separately in inventories. The Dutch NCSC rates the likelihood of exploitation as medium and the potential damage as high.

What should IT administrators do now?

Map out where Java 17.0.6 or newer is running on servers with a recent Intel or AMD processor, prioritising internet-facing systems such as Jira, Confluence, Jenkins and Keycloak. Install the fixed build, restart the Java services and rebuild container images; an update only takes effect after a restart. If your distribution does not have a fix yet, temporarily start Java with -XX:-UseBASE64Intrinsics and test performance beforehand. Watch the logs for spikes in IllegalArgumentException errors from the Base64 decoder and for unexplained JVM crashes.

top ^

Optimizely hacked: websites of KLM, ABN AMRO and ICS Cards served skimmer and iPhone exploits

Vulnerability Supply chain attack via the marketing and A/B testing platform Optimizely Web. Attackers placed malicious code in the Project JavaScript field of customer projects, which was then served via the trusted cdn.optimizely.com. How they obtained write access is unknown
CVE number No CVE assigned
Actively exploited Yes, on 16 November from 04:21 UTC until the major clean-up around 21:00 UTC; 188 snippets from 178 organisations infected, according to Report URI
Patch available Not applicable. Optimizely says it has closed the hole and removed the code. The iOS exploits no longer work from iOS 18.7.3 and iOS 26.3 onwards

How was this exploited?

Scott Helme of Report URI discovered that on 16 November visitors to 178 organisations’ websites received malicious JavaScript via Optimizely. These include KLM, Transavia, ABN AMRO, ICS Cards and Heineken, alongside the BBC, CNN, Zoom, GitLab and IKEA, among others. The code replaced cryptocurrency wallet addresses with the attacker’s, skimmed credit card details from forms from 13:43 UTC, sent cookies and screenshots to the domain 1688360.cc and attempted to infect iPhones with the Coruna and DarkSword exploit kits. According to Dutch security news site Security.nl, the attackers netted around 7,000 dollars in crypto. Report URI had previously seen the same attacker abuse scripts from Kissmetrics, StorifyMe and Minute Media.

What could the impact be?

Card details entered into forms on affected sites on 16 November should be considered compromised, as should session data in cookies without HttpOnly. Payment fields in iframes, such as Stripe’s, were not reachable. How many cards or iPhones were affected is unknown. Of the organisations examined, only 15 per cent blocked the attack with a Content Security Policy.

What should IT administrators do now?

If your clients use Optimizely Web, check all Project JavaScript and experiment code for 1688360.cc, atob( and unknown scripts, review who and what has write access, and rotate API tokens and passwords. Block the domains on Report URI’s list in your DNS filter, including their subdomains. Deploy a CSP with a restrictive connect-src and avoid ‘unsafe-eval’. Update iPhones and iPads to at least iOS 18.7.3 or iOS 26.3.

top ^

N-able N-central: Hotfix 3 now available with seventeen new fixes, legacy APIs closed for good

Vulnerability Seventeen flaws in N-able N-central prior to 2026.4 Hotfix 3, including an authentication bypass on the MMS3 appliance via a predictable passcode, SOAP operations without authorisation checks, path traversal leading to code execution, syslog injection leading to root commands, XXE and SSRF
CVE number CVE-2026-129271 to CVE-2026-129287 inclusive
CVSS score Up to 9.3 (CVE-2026-129272 and CVE-2026-129280, CVSS 4.0)
Actively exploited Not reported for these seventeen flaws. However, the previously published exploit chain 4EverAble2Shell has been exploited since 12 November
Patch available Yes, N-central 2026.4 Hotfix 3 (build 2026.4.1.39). N-able updates hosted servers itself; all Agents and Probes must also be updated to HF3

What’s new?

The Patchwork of 13 November described how a new exploit bypassed N-able’s hotfix. Hotfix 3 has now been released. According to the release notes, it fixes seventeen flaws and removes the configuration option for the legacy APIs, so those APIs are now permanently blocked. It was precisely those legacy APIs that were needed for the 4EverAble2Shell chain, which researcher Sean Bales published on 12 November and which, according to SecurityWeek, was exploited within hours to deploy cryptominers. HF3 also adds agent verification and a device identification requirement.

What could the impact be?

Several of the flaws can be exploited without authentication. An attacker can impersonate an appliance, inject task results and asset data, download clients’ policy information and tamper with the Take Control audit log. With an RMM, that means access to the systems of all your clients. N-able also warns that Agents and Probes that have not been updated are ‘susceptible to impersonation’.

What should IT administrators do now?

If you run N-central on-premises, update the server to build 2026.4.1.39 immediately; a direct upgrade is possible from 2025.4. Then update all Agents and Probes, even if you use the hosted version. Check whether any integrations still rely on the legacy APIs, as these will no longer work after the update. Keep the management console behind a VPN or zero-trust access.

top ^

Puppet Enterprise flaw actively exploited: attackers execute code on the Puppet server

Vulnerability Deserialisation of untrusted data in the report processing of Puppet Server in Perforce Puppet Enterprise 2023.8.0 to 2023.8.13, 2025.0.0 to 2025.12.1 and 2026.0.0
CVE number CVE-2026-115731
CVSS score 9.4 (CVSS 4.0, Puppet)
Actively exploited Yes, added by CISA to the Known Exploited Vulnerabilities catalogue on 17 November
Patch available Yes, since 3 November: Puppet Enterprise 2023.8.13, 2025.12.1 and 2026.0.1

How can the vulnerability be exploited?

A client with an agent certificate signed by Puppet’s certificate authority can submit a report with a manipulated data type via a PUT request and thereby execute code on the Puppet Server, under the service account. If the CA is configured to sign automatically without strong verification, an unauthenticated attacker can also obtain such a certificate, according to the CVE description. CISA reports that the flaw is being actively exploited. The agency has not provided details of the attacks.

What could the impact be?

The Puppet Server determines the configuration of all managed servers. An attacker who takes over that server can use Puppet itself to roll out code across the entire server estate. In the same releases, Perforce also fixed a second critical flaw, an authorisation bypass on certificate names.

What should IT administrators do now?

Update Puppet Enterprise to 2023.8.13, 2025.12.1 or 2026.0.1 and restart Puppet Server. Review the CA’s autosign policy and disable automatic signing if there is no strong verification in place. Restrict network access to Puppet Server to the managed nodes, and check the list of signed certificates for any unknown agents.

top ^

Zyxel patches critical flaw in modems, fibre ONTs and Wi-Fi extenders, also in use at Dutch provider Odido

Vulnerability Unauthenticated out-of-bounds write in libclinkc.so, plus a post-authentication command injection as administrator, in more than thirty Zyxel models: DSL and Ethernet CPE, fibre ONTs and Wi-Fi extenders, including the DX, EX, EMG, VMG, PM and WX series and the EX5601-T1
CVE number CVE-2026-79614 and CVE-2026-86676
CVSS score 9.8 for CVE-2026-79614 (CVSS 3.1, Zyxel)
Actively exploited No, not reported
Patch available Yes, new firmware per model (for example EX5601-T0 5.70(ACDZ.6.4)C0 and DX3300-T0 5.50(ABVY.8.4)C0), available on request from Zyxel support. For provider versions, the update is delivered via the provider

How can the vulnerability be exploited?

According to Zyxel’s advisory, an unauthenticated attacker can use a specially crafted HTTP request to escalate privileges and potentially execute commands on the device. The second flaw requires administrator access and can only be used if the password has already been leaked. Zyxel stresses that WAN access is disabled by default on these devices, so an attacker must reach the device from the local network, or WAN management must have been enabled. The flaws were discovered by Adam Simuntis and Joni Gadd.

What could the impact be?

A compromised modem or ONT gives an attacker a foothold at the edge of the network, where they can intercept or redirect traffic. Security.nl points out that the EX5601-T1 that Odido supplies to its customers is also affected. Zyxel does not list the special provider models in its tables; whether and when Odido and other providers will roll out the update is unknown.

What should IT administrators do now?

Take stock of which Zyxel modems, ONTs and extenders are installed at your SME clients and compare the firmware against the tables in the advisory. For your own devices, request the new firmware from Zyxel, and for provider equipment, ask when the provider will roll out the update. Make sure WAN management is disabled and replace default passwords.

top ^

This article was originally published in Dutch on MSP Business. Read the original →