Introduction
Oracle's November 2026 Critical Security Patch Update fixes an access control flaw in the Security Framework of Oracle Application Development Framework (ADF). The flaw lets an unauthenticated attacker with HTTP access read, create, modify, or delete data that the framework is supposed to protect. CVE-2026-91618 has a CVSS 3.1 score of 9.1 and requires no login and no user interaction, so organizations running ADF 12.2.1.4.0 should patch promptly even though no public exploitation has been confirmed.
Oracle ADF is a Java EE framework within Oracle Fusion Middleware. Oracle positions it as a way to simplify application development through reusable infrastructure services and visual, declarative development. The reviewed sources do not give a deployment count or market share figure for ADF. However, ADF applications run inside Fusion Middleware environments such as WebLogic domains, and the Security Framework sits in the request path that decides who may see and change application data.
Technical Information
Root cause
NVD classifies CVE-2026-91618 as CWE-284, Improper Access Control. MITRE defines this as a product failing to restrict, or incorrectly restricting, an unauthorized actor's access to a resource. CWE-284 is a broad category. It does not name a specific faulty function or technique.
The ADF patch therefore changes how the Security Framework decides whether an unauthenticated HTTP request may read or change protected data. Oracle has not said which classes, filters, or authorization checks were changed, and there is no public commit, diff, or source level changelog. Any claim about the exact code level change should be treated as unverified.
Reading the CVSS vector
The published vector is CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:N:
- Attack Vector: Network. Oracle lists the protocol as HTTP, and its advisory notes that a listed protocol includes its secure variant, so HTTPS is in scope too.
- Attack Complexity: Low. Oracle describes the flaw as easily exploitable.
- Privileges Required: None. No authenticated account is needed.
- User Interaction: None. No victim action is required.
- Scope: Unchanged. The data exposure can still be large. The stated impact can extend to all data accessible through ADF.
- Confidentiality: High. Unauthorized access to critical data, or complete access to all ADF accessible data.
- Integrity: High. Unauthorized creation, deletion, or modification of critical data or all ADF accessible data.
- Availability: None. The published assessment gives no availability impact, so this should not be described as a denial of service issue.
Attack flow based on the published description
Oracle has disclosed the prerequisites and the consequences, but not the mechanics. Within those limits, an attack would look like this:
- The attacker finds an ADF 12.2.1.4.0 application that is reachable over HTTP or HTTPS from their network position.
- The attacker sends requests with no credentials and no session. No victim interaction is involved.
- The ADF Security Framework fails to restrict access correctly (CWE-284), and the request is processed against resources that should require authorization.
- The attacker gains read access to critical or all ADF accessible data, the ability to create, delete, or modify that data, or both.
The public records do not include a vulnerable URL path, HTTP method, request structure, payload, faulty class, or reproducible procedure. The evidence supports an access control analysis but not a reconstruction of the exploit.
What the disclosure does not say
Nothing in the advisory, CVE record, or NVD entry describes operating system command execution, deserialization, or injection. Do not assume any of these, or any particular authentication bypass mechanism, when assessing exposure.
Patch Information
Oracle fixed CVE-2026-91618 in its November 2026 Critical Security Patch Update (CSPU), released on 17 November 2026 as advisory Rev 1, the initial release. As with most Oracle middleware issues, the fix is delivered only as a binary patch through My Oracle Support (MOS).
Risk matrix entry
The Fusion Middleware risk matrix lists the following values:
- Product and component: Oracle Application Development Framework (ADF), Security Framework
- Protocol: HTTP
- Remotely exploitable without authentication: Yes
- CVSS 3.1: 9.1 (
AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:N) - Supported version affected: 12.2.1.4.0
- Notes or footnotes: None
Related ADF fixes in the same update
The ADF section of the November 2026 matrix lists CVE-2026-91618, CVE-2026-91619, and CVE-2026-91670. Oracle ships Fusion Middleware security fixes as product patches rather than one patch per CVE. The ADF patch for this CVE will most likely cover the other two as well, but the advisory does not confirm the exact bundling, so plan and test them as one change. Across all of Fusion Middleware, this CSPU contains 71 new security patches, 53 of which address issues that can be exploited remotely without authentication.
Why the October 2026 CPU is not enough
In 2026 Oracle added a second patch track:
- Critical Patch Updates (CPUs): quarterly and cumulative.
- Critical Security Patch Updates (CSPUs): smaller, targeted releases of high priority fixes, published between CPUs.
Oracle says CSPUs complement the quarterly CPUs. A 12.2.1.4.0 environment that is fully patched to the October 2026 CPU remains vulnerable until the November CSPU patch for ADF is installed. Teams that only patch on the quarterly CPU schedule need to add this release.
Obtaining the patch
Patch numbers and download links are available only inside My Oracle Support, which requires a login. The advisory points to two MOS references:
- Doc ID KA1182: the full list of current and earlier CPU and CSPU patches for Fusion Middleware products.
- Note CPU450: the Fusion Middleware Patch Availability Document linked from the ADF row. Use it to find the ADF patch and its installation instructions.
None of the sources we reviewed publish the patch number. Take it from these MOS documents, not from third party sites.
Support coverage and interim measures
Oracle produces CSPU patches only for releases under Premier or Extended Support. Releases outside those phases are not tested against these vulnerabilities, and Oracle says earlier versions are likely affected too. No backport will be provided for them; Oracle's guidance is to move to a supported release.
Oracle's interim measures, such as blocking the attack protocol or removing unnecessary access, can break functionality and are not long term fixes because they do not correct the underlying problem. Removing access is also of limited value here, since the flaw requires no privileges. The advisory gives no endpoint specific filtering rule and no confirmed WAF signature. Applying the patch is the only real fix.
Verifying the fix
Oracle middleware patches are usually installed with OPatch. After applying the patch, compare the installed inventory against the patch ID listed in CPU450 or KA1182. Do this on every ADF runtime host, including WebLogic domains that host ADF applications.
Detection Methods
As of 18 November 2026 there is no tested signature for CVE-2026-91618. Detection comes down to two things: finding vulnerable ADF deployments, and watching ADF backed applications for unusual unauthenticated activity.
1. Exposure detection (most reliable)
Treat a host as exposed if all of the following are true:
- It runs ADF 12.2.1.4.0.
- The Fusion Middleware fix from the November 2026 CSPU (tracked under MOS Note CPU450) is not applied.
- Its ADF applications can be reached over HTTP.
Because the vector is AV:N/AC:L/PR:N/UI:N, put ADF applications reachable from the internet or other untrusted networks at the top of the list. Commercial scanners (for example Rapid7, which has a database entry for this CVE) and attack surface tools (IONIX offers exposure scanning for it) usually flag this kind of issue by version and patch level. They do not confirm exploitation.
2. Behavioral monitoring
Oracle has not published the root cause, the vulnerable endpoint, or a request pattern, so an exploit specific signature is not possible yet. The impact description still tells you what a successful attack would look like: a requester with no login reading data that ADF should restrict, or creating, deleting, or modifying it.
IONIX recommends reviewing ADF based applications for unexpected access or data changes. CVETodo advises enabling detailed logging and monitoring suspicious HTTP activity aimed at ADF components. In practice:
- Web, proxy, and WebLogic access logs: look for requests with no session or credentials that receive successful responses on resources that should require a login.
- Application or database audit trails: link those requests to writes, changes, or bulk reads.
An unauthenticated request followed by a data change is currently the strongest available sign of exploitation. Without a published request pattern, these are anomaly hunts, not signature matches.
3. Community templates: use with caution
TheHackerWire publishes a "Defender's Toolkit" for this CVE, labeled "Generic detection templates, tune before production." Take that warning seriously:
- Sigma rule (marked experimental): selects web server responses with status 500 or 502. Its condition refers to
1 of indicators*, but no indicators are defined, so it will not work as written. - Splunk hunting query: searches web and firewall indexes for the literal strings "unauthorized creation" or the CVE ID with status 400 or above, then counts results by source IP, URI path, and status. Real attack traffic is unlikely to contain text from the CVE description. Grouping by source IP, URI path, and status is a reasonable starting point for the anomaly hunt above once you substitute your own ADF application paths and authentication signals.
- Nuclei invocation by CVE ID: OffSeq reports that no public Nuclei template is indexed, and CVETodo lists none. The command will most likely match nothing.
- ModSecurity rule: its regex looks for path traversal, script tags,
union select, and JNDI strings. These are generic injection patterns with no stated link to an access control flaw in the ADF Security Framework, so they should not count as coverage for this CVE.
TheHackerWire's triage checklist also suggests searching SIEM and EDR data back to the 17 November 2026 publication date. That suits a CVE with no known indicators, because older logs can be rechecked once details are released.
4. Current state of signatures and IoCs
- CVETodo reports no published Sigma, Suricata, YARA, or Nuclei rules as of 17 November 2026.
- OffSeq records no public exploit code in Metasploit or Exploit DB, no CISA KEV listing, and no attack traffic on its Mirage honeypot sensors.
- TheHackerWire reports no recorded active exploitation.
- No IoCs (IP addresses, payloads, URIs, or user agents) have been published, and no threat actor has been linked to this CVE.
These negative findings are limited to what each source can see. Check vendor IDS and WAF rule feeds and scanner plugin updates again as Oracle or researchers release technical details.
Affected Systems and Versions
- Product: Oracle Application Development Framework (ADF), part of Oracle Fusion Middleware
- Component: Security Framework
- Affected supported version: 12.2.1.4.0
- Vulnerable configuration: ADF 12.2.1.4.0 without the November 2026 CSPU ADF patch, with ADF applications reachable over HTTP or HTTPS. This includes WebLogic domains that host ADF applications.
- Patch level note: environments that are current only to the October 2026 CPU remain vulnerable.
- Version 14.1.2.0.0: this version appears in the advisory's broader ADF product inventory, but the CVE specific row lists only 12.2.1.4.0. Do not label 14.1.2.0.0 as affected by CVE-2026-91618 based on the inventory alone.
- Unsupported releases: Oracle does not test releases outside Premier or Extended Support and states that earlier versions are likely affected. No patch will be provided for them.
Vendor Security History
Oracle describes Oracle Software Security Assurance as its method for building security into product design, development, testing, and maintenance. Its Product Security organization operates under the Chief Security Officer. Oracle's remediation policy uses Critical Patch Updates as the main channel for backported security fixes, with Security Alerts for exceptional cases. In 2026 it added the CSPU track that delivered this fix.
Oracle enterprise products have also been exploited. CVE-2025-61882, a separate vulnerability in Oracle E Business Suite, was the subject of an Oracle Security Alert on 4 October 2025. Oracle described remote exploitation without authentication and potential remote code execution. Google Threat Intelligence later documented an exploitation and extortion campaign against E Business Suite, with activity likely dating back to July 2025 and extortion emails containing addresses associated with the CL0P leak site.
That case is relevant background on threats to Oracle enterprise applications. It is not evidence that CL0P, or the same exploit chain, is involved with this ADF vulnerability, and its indicators should not be applied to CVE-2026-91618.
References
- NVD entry for CVE-2026-91618
- CVE Record for CVE-2026-91618
- Oracle Critical Security Patch Update Advisory, November 2026
- Text Form of Oracle CSPU November 2026 Risk Matrices
- Oracle Security Blog: November 2026 Critical Security Patch Update Released
- My Oracle Support Note CPU450, Fusion Middleware Patch Availability Document (login required)
- Rapid7 Vulnerability Database entry for CVE-2026-91618
- TheHackerWire: CVE-2026-91618 Exploit Status and Fix
- Threat Radar (OffSeq): CVE-2026-91618 Improper Access Control
- IONIX Threat Center: CVE-2026-91618
- CVETodo: CVE-2026-91618
- MITRE CWE-284: Improper Access Control
- CISA Known Exploited Vulnerabilities Catalog
- Oracle Application Development Framework product page
- Oracle Security Alert for CVE-2025-61882
- Google Cloud Threat Intelligence: Oracle E Business Suite Zero Day Exploited in Extortion Campaign
- Oracle Security Vulnerability Remediation Practices
- Oracle Global Product Security



