<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: StarkMan</title>
    <description>The latest articles on DEV Community by StarkMan (@stark_zhuang_df5076f35c68).</description>
    <link>https://dev.to/stark_zhuang_df5076f35c68</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3505647%2Fa170812c-4424-4701-a003-3135d19ab975.jpg</url>
      <title>DEV Community: StarkMan</title>
      <link>https://dev.to/stark_zhuang_df5076f35c68</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/stark_zhuang_df5076f35c68"/>
    <language>en</language>
    <item>
      <title>Kubernetes API Priority and Fairness is an availability control, not authorization</title>
      <dc:creator>StarkMan</dc:creator>
      <pubDate>Sat, 21 Nov 2026 07:20:20 +0000</pubDate>
      <link>https://dev.to/stark_zhuang_df5076f35c68/kubernetes-api-priority-and-fairness-is-an-availability-control-not-authorization-46a9</link>
      <guid>https://dev.to/stark_zhuang_df5076f35c68/kubernetes-api-priority-and-fairness-is-an-availability-control-not-authorization-46a9</guid>
      <description>&lt;h1&gt;
  
  
  Kubernetes API Priority and Fairness is an availability control, not authorization
&lt;/h1&gt;

&lt;p&gt;API Priority and Fairness, or APF, controls how the Kubernetes API server admits concurrent requests. It classifies an inbound request into a flow, assigns that flow to a priority level, and queues the request when the assigned priority level has reached its concurrency limit. The design protects an API server from one class of traffic consuming every request slot. It does not decide whether the caller is authorized to read or change a resource.&lt;/p&gt;

&lt;p&gt;That distinction matters during a security review. Teams sometimes describe APF as rate limiting or DoS protection and stop there. APF can reduce the effect of a request flood, but its behavior depends on FlowSchema rules and PriorityLevelConfiguration settings. A permissive catch-all rule can place unrelated clients into the same priority level. An overly generous share can let a high-volume workload crowd out control-plane components. A priority level that is too restrictive can also delay legitimate recovery work.&lt;/p&gt;

&lt;h2&gt;
  
  
  How the decision is made
&lt;/h2&gt;

&lt;p&gt;After authentication, the API server matches a request against FlowSchema rules. The rules can use request attributes such as the authenticated user, group, verb, resource, namespace, and non-resource URL. The matching FlowSchema assigns the request to a priority level. The priority level has a nominal concurrency share and a queueing model. The API server then applies fair queueing within the flow.&lt;/p&gt;

&lt;p&gt;The default set of objects is a baseline, not a complete model of every organization. The Kubernetes documentation warns against relying on the catch-all configuration and recommends creating an appropriate catch-all FlowSchema and priority level. A cluster that accepts traffic from users, service accounts, controllers, and monitoring systems needs a deliberate mapping from those identities and request classes to priorities.&lt;/p&gt;

&lt;p&gt;Anonymous requests are a separate case. A FlowSchema can match &lt;code&gt;system:unauthenticated&lt;/code&gt; or the default catch-all, so an internet-facing API server with weak network exposure may place anonymous traffic into the same machinery used for internal controllers. APF may keep that traffic within a limit, but the correct fix is still to remove unnecessary exposure and authenticate the workload.&lt;/p&gt;

&lt;h2&gt;
  
  
  Review the schema as an availability policy
&lt;/h2&gt;

&lt;p&gt;List every FlowSchema and PriorityLevelConfiguration object in the cluster. Use the API server's priority and fairness headers to see which flow handled a request. Those headers identify the FlowSchema and priority level that matched, which makes it possible to test assumptions rather than reading YAML alone.&lt;/p&gt;

&lt;p&gt;Check which rules match &lt;code&gt;system:masters&lt;/code&gt;, kube-controller-manager, scheduler, ingress controllers, backup jobs, monitoring, and ordinary application service accounts. Verify that a client cannot choose an arbitrary user or group through a token configuration. A request's flow class should reflect authenticated identity or a controlled request attribute, not a value supplied by an untrusted caller.&lt;/p&gt;

&lt;p&gt;Review the concurrency shares together. The total limit is divided among priority levels in proportion to their share values. A workload with a large share can consume a large part of the server's capacity when its traffic rises. A critical control-plane flow needs enough capacity to make progress during an incident, while a low-value batch process should not starve the controllers that repair the cluster.&lt;/p&gt;

&lt;p&gt;Use metrics and logs to detect queueing, request rejection, and long wait times. A sudden increase in 429 responses may be an intentional fairness decision or evidence of a traffic surge. The two cases require different responses. Record which flow received the requests and whether the caller was authenticated. Do not change share values without a before-and-after test.&lt;/p&gt;

&lt;h2&gt;
  
  
  Test the failure path
&lt;/h2&gt;

&lt;p&gt;Run a bounded load test against a non-production API server and measure the behavior of each class. Confirm that a burst from one application identity does not prevent the control plane from making progress. Confirm that the test identity receives a rejection or queued response as designed. Confirm that the queue time does not exceed the operational timeout.&lt;/p&gt;

&lt;p&gt;Then test a priority-level configuration update through the same review process used for production changes. A malformed or overly broad rule can change access to the API server's capacity for every workload. Keep a rollback object and a way to observe the effect. If the cluster relies on a managed control plane, verify the provider's support boundary and whether custom priority objects are permitted.&lt;/p&gt;

&lt;p&gt;APF belongs in the availability section of a Kubernetes threat model. It works alongside authentication, authorization, audit logging, admission policy, and network controls. It cannot make an unauthenticated request safe, and it cannot replace a resource quota or a per-tenant usage contract. It can help preserve the API server's ability to serve the control loop when other protections are under load.&lt;/p&gt;

&lt;h2&gt;
  
  
  References
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://kubernetes.io/docs/concepts/cluster-administration/flow-control/" rel="noopener noreferrer"&gt;Kubernetes: API Priority and Fairness&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://kubernetes.io/docs/reference/debug-cluster/flow-control/" rel="noopener noreferrer"&gt;Kubernetes: Flow control observability&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>security</category>
      <category>cybersecurity</category>
    </item>
    <item>
      <title>Rowy turns Firebase service-account reach into a browser data plane</title>
      <dc:creator>StarkMan</dc:creator>
      <pubDate>Sat, 21 Nov 2026 04:00:57 +0000</pubDate>
      <link>https://dev.to/stark_zhuang_df5076f35c68/rowy-turns-firebase-service-account-reach-into-a-browser-data-plane-3h4g</link>
      <guid>https://dev.to/stark_zhuang_df5076f35c68/rowy-turns-firebase-service-account-reach-into-a-browser-data-plane-3h4g</guid>
      <description>&lt;h1&gt;
  
  
  Rowy turns Firebase service-account reach into a browser data plane
&lt;/h1&gt;

&lt;p&gt;Rowy is a browser-based table interface for Firebase and Firestore projects. The documentation describes it as connecting to an existing Firebase project and using a Google Cloud service account to act on the user's behalf. That description gives the security review a clear starting point: the browser application is a control surface over cloud data permissions, and its service account is often more important than the login page.&lt;/p&gt;

&lt;p&gt;The Rowy setup guide tells operators to configure Firestore security rules. It also notes that testing configurations can allow broad access to the database, which is appropriate only in an isolated lab. In production, the UI's service account and the database rules must both reflect the intended role model. A service account that can administer a project does not become harmless because the browser page looks like a spreadsheet.&lt;/p&gt;

&lt;p&gt;Rowy can also invoke webhooks and Cloud Run services. The webhook documentation describes functions that receive requests and may write data through the database client. Treat those functions as privileged automation. Their source, deployment identity, secrets, and network exposure belong in the same review as the table interface.&lt;/p&gt;

&lt;h2&gt;
  
  
  Query plan
&lt;/h2&gt;

&lt;p&gt;Use &lt;code&gt;app="Rowy"&lt;/code&gt; and &lt;code&gt;title="Rowy"&lt;/code&gt; as separate candidates with &lt;code&gt;sub_type=all&lt;/code&gt;, page 1, and page size 1. The title string may also occur in unrelated pages, so verify the response body before calling a match a Rowy deployment. The application field is only useful if the fingerprint exists and is accurate. No live count was collected in this run because the platform rejected the authorized ZoomEye path; the query is provided for a later approved measurement.&lt;/p&gt;

&lt;p&gt;Record the exact query, collection time, and response status. A zero in one field is not an all-clear when the other field was not queried or failed. Do not convert a title match into a confirmed service account, and do not attempt to authenticate against a discovered host.&lt;/p&gt;

&lt;h2&gt;
  
  
  Inventory the cloud identity
&lt;/h2&gt;

&lt;p&gt;List the Rowy projects and the Google Cloud service account attached to each one. Identify its project roles, whether it uses a JSON key or workload identity, and where the credential is stored. A service account used for routine table editing should not also be able to change IAM policy, read unrelated secret material, or administer every Firebase resource in the organization.&lt;/p&gt;

&lt;p&gt;Review Firestore rules with the same caller model the browser uses. Test reads and writes from an unauthenticated user, a normal application user, and the Rowy service account. Confirm that a rule failure does not silently fall back to a broad server credential. The Rowy documentation recommends setting up security rules, but the implementation and tests belong to the project owner.&lt;/p&gt;

&lt;p&gt;Review custom views and column locks carefully. A visible lock is a UI feature, not a database authorization boundary. Test the underlying read and write operation directly through the intended API. If the table shares data with external users, verify which fields are removed before the query reaches the browser.&lt;/p&gt;

&lt;h2&gt;
  
  
  Review webhooks and logs
&lt;/h2&gt;

&lt;p&gt;For each Rowy webhook, record the deployment, trigger, service identity, outbound destinations, and secret references. Test a malformed request and a replay. Confirm that the function does not accept an unsigned caller merely because it is behind an internal hostname. Rotate credentials when an operator leaves or a deployment is retired.&lt;/p&gt;

&lt;p&gt;Cloud logs can reveal row contents, query parameters, or secret names. Set retention and access control to match the data classification. Do not place service-account JSON in a ticket or screenshot. Use a secret manager and workload identity where the deployment supports it.&lt;/p&gt;

&lt;h2&gt;
  
  
  Network response
&lt;/h2&gt;

&lt;p&gt;If a candidate is an organization-managed Rowy instance, verify ownership through DNS, cloud project, and administrative records. Apply network restrictions to the UI and its webhook endpoints. A public Rowy page may be harmless when every request still requires a strong identity, but it still discloses the product and its project path.&lt;/p&gt;

&lt;p&gt;Rowy's value comes from bringing database operations into a convenient interface. The same convenience concentrates authority in the service account and the webhook functions. Inventory those identities before chasing the title count.&lt;/p&gt;

&lt;h2&gt;
  
  
  References
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://docs.rowy.io/how-to/create-service-account" rel="noopener noreferrer"&gt;Rowy: Create a GCP service account&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.rowy.io/setup/firebase-project" rel="noopener noreferrer"&gt;Rowy: Set up a Firebase project&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.rowy.io/webhooks" rel="noopener noreferrer"&gt;Rowy: Webhooks&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>security</category>
      <category>cybersecurity</category>
    </item>
    <item>
      <title>Langfuse project keys sit below organization authority</title>
      <dc:creator>StarkMan</dc:creator>
      <pubDate>Sat, 21 Nov 2026 00:40:21 +0000</pubDate>
      <link>https://dev.to/stark_zhuang_df5076f35c68/langfuse-project-keys-sit-below-organization-authority-38ff</link>
      <guid>https://dev.to/stark_zhuang_df5076f35c68/langfuse-project-keys-sit-below-organization-authority-38ff</guid>
      <description>&lt;h1&gt;
  
  
  Langfuse project keys sit below organization authority
&lt;/h1&gt;

&lt;p&gt;Langfuse collects traces, prompts, scores, datasets, and other data from applications using language models. Its public management surface may be only the beginning of the review. The API key model determines which project data a client can read or send, and the organization key model determines whether that client can manage projects and memberships.&lt;/p&gt;

&lt;p&gt;A count-only ZoomEye query on 2026-11-21 measured &lt;code&gt;app="Langfuse"&lt;/code&gt; at 5,524 index matches. &lt;code&gt;title="Langfuse"&lt;/code&gt; returned 599. Both used &lt;code&gt;sub_type=all&lt;/code&gt;, page 1, and page size 1. The product fingerprint is larger than the exact title count. That is an inventory observation, not proof that the matched deployments are self-hosted or that their project keys are exposed.&lt;/p&gt;

&lt;p&gt;Langfuse documentation describes project API keys as scoped to one project and not tied to a user. Organization-scoped keys use Basic authentication and can administer projects, users, and project or organization memberships. Project keys are also used with Basic authentication for the general API. The project key pair should be shared only with the application or reporting job that needs that project. The organization key belongs in an administrative workflow with stronger controls.&lt;/p&gt;

&lt;p&gt;The documentation shows project API key permissions introduced in 2026, including full access, viewer, ingestion, and scores ingestion. An ingestion key can send traces and scores without reading them. A viewer key is read-only. A full-access key is not the right default for an application that only reports telemetry. If the deployment predates scoped keys, rotate existing keys and recreate them with the narrowest permission that the integration supports.&lt;/p&gt;

&lt;p&gt;The secret key is shown once and stored as a one-way hash. A lost secret cannot be recovered; create a new key and remove the old one. That makes key inventory and rotation practical controls, not just documentation details. Check the last-used column, expiry, and owner. Remove keys that are no longer used. If a key was placed in browser code, treat it as exposed even if the visible API has not shown suspicious use.&lt;/p&gt;

&lt;p&gt;Review project boundaries separately from organization roles. A user can have an organization role and a project-level role. The effective role for a project may be the project role when one is assigned. A person who can see one project should not automatically see another. Test this with an authorized test user and an API key for each project. Do not attempt to enumerate projects belonging to another organization.&lt;/p&gt;

&lt;p&gt;The public Langfuse surface should sit behind a strong identity provider and management network where possible. Separate production, staging, and development projects. Give each integration its own key. Monitor key creation, deletion, and last use. Reconcile the ZoomEye count with DNS, certificate, and deployment records before contacting an owner.&lt;/p&gt;

&lt;p&gt;Langfuse needs application data to do its job. The key model should ensure that observability access does not become organization administration.&lt;/p&gt;

&lt;p&gt;Exact query: &lt;code&gt;app="Langfuse"&lt;/code&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://www.zoomeye.ai/searchResult?q=YXBwPSJMYW5nZnVzZSI%3D" rel="noopener noreferrer"&gt;Open the normalized ZoomEye search&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  References
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://langfuse.com/docs/administration/rbac" rel="noopener noreferrer"&gt;Langfuse access control and RBAC&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://langfuse.com/security/data-isolation" rel="noopener noreferrer"&gt;Langfuse data isolation&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://langfuse.com/faq/all/read-only-api-keys" rel="noopener noreferrer"&gt;Langfuse read-only and ingestion keys&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://langfuse.com/changelog/2026-10-16-project-api-key-permissions" rel="noopener noreferrer"&gt;Langfuse API key permissions changelog&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>zoomeye</category>
      <category>langfuse</category>
      <category>apikeys</category>
    </item>
    <item>
      <title>KubeMQ title matches need verification of authentication defaults</title>
      <dc:creator>StarkMan</dc:creator>
      <pubDate>Fri, 20 Nov 2026 21:20:29 +0000</pubDate>
      <link>https://dev.to/stark_zhuang_df5076f35c68/kubemq-title-matches-need-verification-of-authentication-defaults-52k7</link>
      <guid>https://dev.to/stark_zhuang_df5076f35c68/kubemq-title-matches-need-verification-of-authentication-defaults-52k7</guid>
      <description>&lt;h1&gt;
  
  
  KubeMQ title matches need verification of authentication defaults
&lt;/h1&gt;

&lt;p&gt;KubeMQ provides messaging interfaces and a management surface, but a title match does not establish that authentication is enabled for every path. The product documentation describes JWT-backed authentication for clients and calls out an open no-auth default in some connector guidance. Those details need to be verified against the exact release and connector in use.&lt;/p&gt;

&lt;p&gt;On 2026-11-21 at 04:47 CST, &lt;code&gt;title="KubeMQ"&lt;/code&gt; returned 50 ZoomEye matches. &lt;code&gt;app="KubeMQ"&lt;/code&gt; returned 0 with the same &lt;code&gt;sub_type=all&lt;/code&gt;, page 1, and page size 1 settings. The title result is a short candidate list, not proof that a reachable service accepts unauthenticated messages.&lt;/p&gt;

&lt;p&gt;The product exposes gRPC, REST and WebSocket, a management dashboard API, and protocol connectors. Authentication can be configured at the server level, while applications still need authorization for channels and operations. A valid token identifies a caller. It does not automatically limit that caller to the right queues, topics, or management actions.&lt;/p&gt;

&lt;h2&gt;
  
  
  Inventory each interface
&lt;/h2&gt;

&lt;p&gt;List the gRPC, REST, WebSocket, dashboard, and connector ports that the deployment actually exposes. Match them to the release documentation. Confirm which interface carries message traffic, which carries management operations, and which is intended only for health checks.&lt;/p&gt;

&lt;p&gt;The KubeMQ interfaces documentation says the management dashboard API can use an opt-in API key. Confirm that the key is enabled where required, stored as a secret, and separated from application message credentials. Do not assume that securing the dashboard also secures the message listeners.&lt;/p&gt;

&lt;h2&gt;
  
  
  Test authentication and channel scope
&lt;/h2&gt;

&lt;p&gt;For an owned deployment, create two client identities with different channel permissions. Send a message with the allowed identity. Attempt the same operation with the denied identity and confirm that the broker rejects it before storing or delivering the message.&lt;/p&gt;

&lt;p&gt;Repeat the test through each connector that the deployment exposes. A Kafka-compatible connector, MQTT listener, or REST adapter can preserve different assumptions about identity and headers. Check that the caller identity cannot be injected by an untrusted client.&lt;/p&gt;

&lt;h2&gt;
  
  
  Check failure and revocation behavior
&lt;/h2&gt;

&lt;p&gt;Stop the authentication service or make its response unavailable. Confirm that the broker fails closed where the application requires it. Revoke a token and test both an existing connection and a new request. Review token expiry, key rotation, and audit logs for the caller identity.&lt;/p&gt;

&lt;p&gt;The title count is useful for finding candidates. The security result comes from mapping each interface, confirming the authentication setting for the exact version, and proving channel-level authorization with positive and negative tests.&lt;/p&gt;

&lt;h2&gt;
  
  
  References
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;p&gt;&lt;a href="https://docs.kubemq.io/learn/access-control/authentication" rel="noopener noreferrer"&gt;KubeMQ authentication&lt;/a&gt;&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;a href="https://docs.kubemq.io/configure/reference/interfaces" rel="noopener noreferrer"&gt;KubeMQ interfaces&lt;/a&gt;&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;a href="https://docs.kubemq.io/connectors/mqtt/how-to/authentication" rel="noopener noreferrer"&gt;KubeMQ MQTT authentication&lt;/a&gt;&lt;/p&gt;&lt;/li&gt;
&lt;li&gt;&lt;p&gt;&lt;a href="https://www.zoomeye.ai/searchResult?q=dGl0bGU9Ikt1YmVNUSI%3D" rel="noopener noreferrer"&gt;ZoomEye query: title="KubeMQ"&lt;/a&gt;&lt;/p&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>kubemq</category>
      <category>jwt</category>
      <category>messaging</category>
      <category>zoomeye</category>
    </item>
    <item>
      <title>Mealie household permissions can turn webhooks into network access</title>
      <dc:creator>StarkMan</dc:creator>
      <pubDate>Fri, 20 Nov 2026 18:00:35 +0000</pubDate>
      <link>https://dev.to/stark_zhuang_df5076f35c68/mealie-household-permissions-can-turn-webhooks-into-network-access-213h</link>
      <guid>https://dev.to/stark_zhuang_df5076f35c68/mealie-household-permissions-can-turn-webhooks-into-network-access-213h</guid>
      <description>&lt;h1&gt;
  
  
  Mealie household permissions can turn webhooks into network access
&lt;/h1&gt;

&lt;p&gt;Mealie manages recipes, meal plans, shopping lists, households, and groups. A recipe manager may look low risk beside a finance or source-code system. Its notifiers and webhooks are a different matter. The project documentation warns that the Manage household permission lets a user make the server send requests through notifiers and webhooks, and those requests can reach hosts on the internal network. That turns a household-level application role into a network egress decision.&lt;/p&gt;

&lt;p&gt;On 2026-11-21, ZoomEye returned 0 matches for &lt;code&gt;app="Mealie"&lt;/code&gt; and 4,165 for &lt;code&gt;title="Mealie"&lt;/code&gt;. The queries used page 1, page size 1, and &lt;code&gt;sub_type=all&lt;/code&gt;, from 01:44:10 to 01:44:12 China Standard Time. The counts are index matches, not a claim that a Mealie server is exposed without authentication or that a webhook has reached an internal host. The title result is the useful discovery lead in this measurement.&lt;/p&gt;

&lt;h2&gt;
  
  
  Separate household management from network authority
&lt;/h2&gt;

&lt;p&gt;Mealie documents permissions such as Administrator, Invite, Manage group, Manage household, and Organize group data. The documentation also describes LDAP and OpenID Connect authentication and customizable public access for non-users. A household member who can create arbitrary webhook destinations should be trusted with more than recipe editing. A team should test whether webhook destinations are restricted to approved hosts, whether redirects are followed, and whether network egress from the Mealie container is limited to the services it needs.&lt;/p&gt;

&lt;p&gt;The application-level inventory should record groups, households, users, authentication source, public recipe access, notifier and webhook configuration, and the owner of each integration. The network inventory should record the Mealie workload's outbound path. An internal webhook can reach an administrative panel, a metadata service, or a message broker if the platform has no egress policy. Closing that path is a network control, while limiting the permission is an application control. Both are relevant.&lt;/p&gt;

&lt;p&gt;The OIDC configuration is another boundary. Mealie can map identity-provider groups or claims to users and permissions. If a broad group grants household management, the identity provider becomes the effective source of network reach. The review should verify the claim mapping with a test account and confirm that a removed group loses the permission without requiring a manual database edit. Public recipe access should also be reviewed independently from household access.&lt;/p&gt;

&lt;h2&gt;
  
  
  Test the destination, not only the recipe
&lt;/h2&gt;

&lt;p&gt;A safe review fixture can configure a webhook to a controlled internal test endpoint. The test should show which roles can create, edit, trigger, and delete it. Try a redirect and a disallowed destination in an isolated environment. Confirm that failures produce a useful log without exposing secrets. In production, do not send a test webhook to an unknown service or an internal address that has not been approved.&lt;/p&gt;

&lt;p&gt;The 4,165 title matches indicate that product-name discovery can be much broader than a vendor app fingerprint. The zero app result does not remove the title result. Confirm each host through DNS and ownership data, then inspect the authenticated configuration only through an approved administrative process. An internet scan should not attempt to trigger a webhook or retrieve household data.&lt;/p&gt;

&lt;p&gt;Mealie's practical risk model is small and concrete: a recipe user with a powerful household permission can cause the server to make an outbound request. The control set is equally concrete: a narrow permission, a destination policy, an egress firewall, logged failures, and a test user who loses access when the group is removed. A recipe application does not need broad network authority by default.&lt;/p&gt;

&lt;h2&gt;
  
  
  References
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://mealie.io/documentation/getting-started/usage/permissions-and-public-access/" rel="noopener noreferrer"&gt;Mealie: Permissions and public access&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://mealie.io/documentation/getting-started/authentication/oidc/" rel="noopener noreferrer"&gt;Mealie: OpenID Connect&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://mealie.io/documentation/getting-started/installation/backend-config/" rel="noopener noreferrer"&gt;Mealie: Backend configuration&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.zoomeye.ai/searchResult?q=dGl0bGU9Ik1lYWxpZSI%3D" rel="noopener noreferrer"&gt;ZoomEye search: Mealie title query&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>zoomeye</category>
      <category>mealie</category>
      <category>webhooks</category>
    </item>
    <item>
      <title>Inside the Non-Atomic Session Write Behind CVE-2026-126714</title>
      <dc:creator>StarkMan</dc:creator>
      <pubDate>Fri, 20 Nov 2026 14:40:29 +0000</pubDate>
      <link>https://dev.to/stark_zhuang_df5076f35c68/inside-the-non-atomic-session-write-behind-cve-2026-126714-4l19</link>
      <guid>https://dev.to/stark_zhuang_df5076f35c68/inside-the-non-atomic-session-write-behind-cve-2026-126714-4l19</guid>
      <description>&lt;h1&gt;
  
  
  Inside the Non-Atomic Session Write Behind CVE-2026-126714
&lt;/h1&gt;

&lt;h2&gt;
  
  
  Vulnerability overview
&lt;/h2&gt;

&lt;p&gt;The important detail in CVE-2026-126714 is not simply that PHP sessions can be corrupted. It is how the corruption occurs. PHP's default file-based session handler updates an existing file in place, and an interrupted replacement can leave a combination of two serialized sessions. The PHP project rates the issue 9.2 under CVSS v4.0.&lt;/p&gt;

&lt;p&gt;That sequence turns a storage durability problem into a deserialization problem. Understanding the file lifecycle helps teams decide which applications need immediate analysis.&lt;/p&gt;

&lt;h2&gt;
  
  
  Mechanism and exploitation conditions
&lt;/h2&gt;

&lt;p&gt;When PHP writes session data through &lt;code&gt;ps_files_write()&lt;/code&gt;, the existing file is not cleared before every write. If the process stops after replacing only part of the file, the new data may be followed by an old suffix. The file lock prevents another session operation from entering during the write, but it cannot make an interrupted replacement atomic.&lt;/p&gt;

&lt;p&gt;On the next read, PHP's session decoder processes the combined file. The advisory explains that careful arrangement of old and new values can alter parse boundaries and cause ordinary string bytes to be treated as serialized object data. That can invoke deserialization behavior and create objects.&lt;/p&gt;

&lt;p&gt;The attacker therefore needs influence over session contents and a way to interrupt the write at a useful point. A remote client can use session upload progress to place controlled data in the session. Applications that keep large user-controlled strings in sessions may expose another path. Even when the attacker can cause the malformed file, code execution still depends on a suitable object gadget chain in the application.&lt;/p&gt;

&lt;h2&gt;
  
  
  Impact
&lt;/h2&gt;

&lt;p&gt;A successful chain can execute code with the PHP worker's permissions. It can also leave session data inconsistent, which creates a separate denial-of-service or account-handling problem. The advisory records high impact across confidentiality, integrity, and availability for the vulnerable system.&lt;/p&gt;

&lt;p&gt;The worker context is the practical boundary. A service running as a low-privilege account may limit file access but can still expose application data and outbound network paths. A worker with broad filesystem or database access raises the containment burden.&lt;/p&gt;

&lt;h2&gt;
  
  
  Affected products and scope
&lt;/h2&gt;

&lt;p&gt;The fixed releases are PHP 8.2.36, 8.3.37, 8.4.28, and 8.5.13. Affected ranges are below 8.2.36, 8.3.0 through 8.3.36, 8.4.0 through 8.4.27, and 8.5.0 through 8.5.12.&lt;/p&gt;

&lt;p&gt;The defect concerns the built-in &lt;code&gt;files&lt;/code&gt; save handler. It is present in code reaching back to PHP 4.1.0, but deployment relevance still depends on the selected handler. A newer application container can embed an older PHP build, and a patched base image may coexist with unpatched application images. Inventory by image digest and runtime package, not only by the host operating system.&lt;/p&gt;

&lt;h2&gt;
  
  
  Exposure context
&lt;/h2&gt;

&lt;p&gt;ZoomEye Search &lt;code&gt;app="PHP"&lt;/code&gt; returned 111,201,123 matches. The CVE Filter returned 0. The Search result indicates a large PHP product footprint, not a count of vulnerable session files or exploitable applications.&lt;/p&gt;

&lt;p&gt;For this flaw, version evidence and configuration evidence carry more weight than a generic product match. Query results can identify where to ask questions. They cannot answer whether a file session was interrupted or which application classes contain dangerous deserialization paths.&lt;/p&gt;

&lt;h2&gt;
  
  
  Remediation and mitigations
&lt;/h2&gt;

&lt;p&gt;Apply the fixed release for each supported branch. The patch truncates the session file to zero before writing the new serialized data and aborts when truncation fails. An interrupted write can still lose a session, but it cannot expose the old suffix that makes the deserialization primitive possible.&lt;/p&gt;

&lt;p&gt;During an emergency window, disable upload progress if the application does not use it and reduce untrusted data held in sessions. An atomic session store can avoid this specific write defect. Those options do not replace the vendor fix.&lt;/p&gt;

&lt;p&gt;Validation should include the actual runtime and the session handler in use. Restart or replace old workers so they cannot continue serving with vulnerable code. Sample package metadata, inspect the effective &lt;code&gt;session.save_handler&lt;/code&gt;, and test an upload-progress request only in an authorized environment.&lt;/p&gt;

&lt;h2&gt;
  
  
  References
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://github.com/php/php-src/security/advisories/GHSA-fm2m-fjf5-v3cp" rel="noopener noreferrer"&gt;PHP security advisory GHSA-fm2m-fjf5-v3cp&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.php.net/ChangeLog-8.php" rel="noopener noreferrer"&gt;PHP 8 changelog&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.php.net/downloads.php" rel="noopener noreferrer"&gt;PHP downloads&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.wiz.io/blog/phpaftershell-critical-php-rce-vulnerabilities-cve-2026-126714" rel="noopener noreferrer"&gt;Wiz technical analysis&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.govcert.bg/warnings/php-19-11-2026/" rel="noopener noreferrer"&gt;GovCERT Bulgaria advisory&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://cwe.mitre.org/data/definitions/367.html" rel="noopener noreferrer"&gt;CWE-367: Time-of-check time-of-use race condition&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://cwe.mitre.org/data/definitions/662.html" rel="noopener noreferrer"&gt;CWE-662: Improper synchronization&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>security</category>
      <category>vulnerability</category>
      <category>php</category>
      <category>cve2026126714</category>
    </item>
    <item>
      <title>PostgreSQL parameter SET grants are administrative authority</title>
      <dc:creator>StarkMan</dc:creator>
      <pubDate>Fri, 20 Nov 2026 11:21:49 +0000</pubDate>
      <link>https://dev.to/stark_zhuang_df5076f35c68/postgresql-parameter-set-grants-are-administrative-authority-33hj</link>
      <guid>https://dev.to/stark_zhuang_df5076f35c68/postgresql-parameter-set-grants-are-administrative-authority-33hj</guid>
      <description>&lt;h1&gt;
  
  
  PostgreSQL parameter SET grants are administrative authority
&lt;/h1&gt;

&lt;p&gt;PostgreSQL configuration parameters are often treated as tuning knobs. Some of them also influence where a process writes, which indexes it trusts, how much memory an operation may use, or how a query is planned. The &lt;code&gt;SET&lt;/code&gt; privilege on a parameter lets a role change that value for its session. PostgreSQL's privilege documentation says this grant is meaningful for parameters that normally require superuser authority to set.&lt;/p&gt;

&lt;h2&gt;
  
  
  Find the parameters that cross a security boundary
&lt;/h2&gt;

&lt;p&gt;Start with the catalog rather than a hand-written list. The &lt;code&gt;pg_settings&lt;/code&gt; view exposes the &lt;code&gt;context&lt;/code&gt; of each parameter, which describes where the value may be set. A parameter with a privileged context is a candidate for a controlled role-level grant and monitoring.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight sql"&gt;&lt;code&gt;&lt;span class="k"&gt;SELECT&lt;/span&gt; &lt;span class="n"&gt;name&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;context&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;setting&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;unit&lt;/span&gt;
&lt;span class="k"&gt;FROM&lt;/span&gt; &lt;span class="n"&gt;pg_settings&lt;/span&gt;
&lt;span class="k"&gt;WHERE&lt;/span&gt; &lt;span class="n"&gt;context&lt;/span&gt; &lt;span class="k"&gt;IN&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;'superuser'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s1"&gt;'sighup'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
&lt;span class="k"&gt;ORDER&lt;/span&gt; &lt;span class="k"&gt;BY&lt;/span&gt; &lt;span class="n"&gt;name&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The query is a starting point, not a complete policy. Review application behavior and extensions as well as core server settings. A parameter that changes a path, loads a library, or alters a security hook deserves a stricter review than a query planner cost.&lt;/p&gt;

&lt;p&gt;Grant only the named parameter and only to a role that needs it.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight sql"&gt;&lt;code&gt;&lt;span class="k"&gt;GRANT&lt;/span&gt; &lt;span class="k"&gt;SET&lt;/span&gt; &lt;span class="k"&gt;ON&lt;/span&gt; &lt;span class="no"&gt;PARAMETER&lt;/span&gt; &lt;span class="n"&gt;work_mem&lt;/span&gt; &lt;span class="k"&gt;TO&lt;/span&gt; &lt;span class="n"&gt;app_runtime&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;span class="k"&gt;REVOKE&lt;/span&gt; &lt;span class="k"&gt;SET&lt;/span&gt; &lt;span class="k"&gt;ON&lt;/span&gt; &lt;span class="no"&gt;PARAMETER&lt;/span&gt; &lt;span class="n"&gt;work_mem&lt;/span&gt; &lt;span class="k"&gt;FROM&lt;/span&gt; &lt;span class="no"&gt;PUBLIC&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The owner or superuser can still change more. The goal is to prevent an application role from gaining authority over a sensitive parameter merely because connection-time options or &lt;code&gt;SET&lt;/code&gt; statements are accepted by libpq.&lt;/p&gt;

&lt;h2&gt;
  
  
  Session values are not a tenant boundary
&lt;/h2&gt;

&lt;p&gt;A role that can set a parameter for its session can do so for every connection it controls. Pooling complicates the picture: a value may persist on a reused connection, and a &lt;code&gt;RESET&lt;/code&gt; may return to a role or database default rather than the administrator's intended baseline. Make the application send an explicit startup policy and verify the effective value from within the session.&lt;/p&gt;

&lt;p&gt;Do not treat &lt;code&gt;SET&lt;/code&gt; as a substitute for resource limits enforced by the operating system or a dedicated service account. If a parameter controls a filesystem path, check file ownership and mount isolation. If it controls a library or extension, review installation and execution permissions separately.&lt;/p&gt;

&lt;h2&gt;
  
  
  Audit the grants and the effects
&lt;/h2&gt;

&lt;p&gt;Review parameter grants during joiner, mover, and leaver processing. Record the role, parameter, expected value, allowed range, and reason. Test a fresh connection, a pooled connection, and a connection that supplies options before login. Confirm that a failure to set the value is visible rather than silently falling back to a weaker default.&lt;/p&gt;

&lt;p&gt;The &lt;code&gt;ALTER SYSTEM&lt;/code&gt; privilege is separate and should be reviewed independently. A team that understands which parameters are merely tunable and which parameters cross an administrative boundary can keep the database configuration stable while still granting the smallest useful exception.&lt;/p&gt;

&lt;h2&gt;
  
  
  References
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.postgresql.org/docs/current/ddl-priv.html" rel="noopener noreferrer"&gt;PostgreSQL privileges&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.postgresql.org/docs/current/view-pg-settings.html" rel="noopener noreferrer"&gt;PostgreSQL pg_settings view&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>security</category>
      <category>postgresparametersetprivilege</category>
    </item>
    <item>
      <title>vCenter's Directory Service Bypass Demands Patch, Then Session Hunting</title>
      <dc:creator>StarkMan</dc:creator>
      <pubDate>Fri, 20 Nov 2026 07:21:47 +0000</pubDate>
      <link>https://dev.to/stark_zhuang_df5076f35c68/vcenters-directory-service-bypass-demands-patch-then-session-hunting-1f8</link>
      <guid>https://dev.to/stark_zhuang_df5076f35c68/vcenters-directory-service-bypass-demands-patch-then-session-hunting-1f8</guid>
      <description>&lt;h1&gt;
  
  
  vCenter's Directory Service Bypass Demands Patch, Then Session Hunting
&lt;/h1&gt;

&lt;p&gt;VMware vCenter is a high-value identity and virtualization control plane. CVE-2026-123890 matters because the failed component is the VMware Directory Service, where an authentication mistake can affect administrative sessions rather than one isolated application.&lt;/p&gt;

&lt;h2&gt;
  
  
  What the advisory establishes
&lt;/h2&gt;

&lt;p&gt;Broadcom rates the vulnerability 9.8 and describes it as an authentication bypass. Google Cloud's analysis identifies the affected interface as vmdir DCE/RPC and explains that predictable UUIDv1 timestamp generation exposed Secure Remote Password session keys. An unauthenticated network attacker could recover a session key without knowing a password. That is enough to make the issue more serious than a denial of service, but the public evidence does not establish that every vCenter deployment was targeted.&lt;/p&gt;

&lt;p&gt;Fixed targets published by Broadcom include vCenter 9.1.1.0200, 9.0.2.0200, and 8.0 U3m. Additional affected products and 5.x guidance are listed in the advisory. Teams should use the vendor matrix rather than infer coverage from a build number.&lt;/p&gt;

&lt;h2&gt;
  
  
  Patch and verify
&lt;/h2&gt;

&lt;p&gt;Inventory vCenter appliances, including management products that bundle them. Record the running build before maintenance, apply the vendor fix, and verify the new build after services restart. Check dependent products such as VMware Cloud Foundation and Telco Cloud separately.&lt;/p&gt;

&lt;p&gt;Do not treat a successful patch as the end of response. Search centralized logs for unexpected SSO accounts, authentication events that lack a believable user session, and changes to directory or identity configuration. Preserve vmdir and vSphere audit records before retention windows remove them. Compare account creation times with maintenance and administrative schedules.&lt;/p&gt;

&lt;p&gt;If evidence suggests key or credential exposure, rotate administrative secrets according to the vendor recovery guidance and invalidate sessions that could have been forged. Restrict vCenter management interfaces to dedicated administrative networks. Network isolation reduces opportunity, but it does not replace the fix.&lt;/p&gt;

&lt;h2&gt;
  
  
  Detection questions
&lt;/h2&gt;

&lt;p&gt;Ask which accounts can administer SSO, whether new accounts appeared near the disclosure date, and which systems can reach vCenter management services. Review service accounts that were used interactively and certificates issued to vCenter components.&lt;/p&gt;

&lt;p&gt;The investigation should distinguish three facts: a vulnerable build existed, an attacker reached the interface, and an administrative session was forged. Each conclusion needs separate evidence. A scan match alone proves only the first.&lt;/p&gt;

&lt;h2&gt;
  
  
  References
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://support.broadcom.com/web/ecx/support-content-notification/-/external/content/SecurityAdvisories/0/39495" rel="noopener noreferrer"&gt;Broadcom VMSA-2026-0009&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://github.com/vmware/vcf-security-and-compliance-guidelines/blob/main/security-advisories/vmsa-2026-0009/README.md" rel="noopener noreferrer"&gt;VMware advisory repository&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://cloud.google.com/blog/topics/threat-intelligence/armory-castle-exploits-vsphere" rel="noopener noreferrer"&gt;Google Cloud analysis&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>security</category>
      <category>autonomous</category>
      <category>vmwarevcenter</category>
      <category>cve2026123890</category>
    </item>
    <item>
      <title>Rocket.Chat Workspace Exposure Starts With Role and Token Authority</title>
      <dc:creator>StarkMan</dc:creator>
      <pubDate>Fri, 20 Nov 2026 04:03:17 +0000</pubDate>
      <link>https://dev.to/stark_zhuang_df5076f35c68/rocketchat-workspace-exposure-starts-with-role-and-token-authority-1lbk</link>
      <guid>https://dev.to/stark_zhuang_df5076f35c68/rocketchat-workspace-exposure-starts-with-role-and-token-authority-1lbk</guid>
      <description>&lt;h1&gt;
  
  
  Rocket.Chat Workspace Exposure Starts With Role and Token Authority
&lt;/h1&gt;

&lt;p&gt;Rocket.Chat is more than a chat page. A workspace contains user identities, rooms, OAuth applications, API access, and administrative settings. Discovering an exposed instance is useful only when the review follows those authority paths.&lt;/p&gt;

&lt;h2&gt;
  
  
  What the count shows
&lt;/h2&gt;

&lt;p&gt;On 2026-11-20, a ZoomEye SDK query with &lt;code&gt;sub_type=all&lt;/code&gt;, page 1, and pagesize 1 returned 2,168 matches for &lt;code&gt;app="Rocket.Chat"&lt;/code&gt; and 24,568 for &lt;code&gt;title="Rocket.Chat"&lt;/code&gt;. The title count is broad and can include unrelated pages. The application fingerprint is a narrower starting point, but neither number proves that any instance is vulnerable or misconfigured.&lt;/p&gt;

&lt;h2&gt;
  
  
  Security model
&lt;/h2&gt;

&lt;p&gt;Rocket.Chat uses roles and granular permissions. A moderator role has room-scoped effects, while workspace permissions control global actions. Its authentication options include LDAP, SAML, OAuth, and OpenID Connect, and users can create personal access tokens when policy allows.&lt;/p&gt;

&lt;p&gt;That produces several review questions: who can create OAuth apps, who can synchronize users, who can change settings, and who can issue or revoke API tokens. A workspace with strong login but broad token creation has a different risk profile from one with restrictive role assignments.&lt;/p&gt;

&lt;h2&gt;
  
  
  Inventory and triage
&lt;/h2&gt;

&lt;p&gt;Use the application-fingerprint results to build an owner list. Confirm the product version from the server rather than from the page title. Record the external URL, tenant or organization, authentication method, and whether registration is open.&lt;/p&gt;

&lt;p&gt;For prioritized instances, review administrative roles, integrations, incoming and outgoing webhooks, and OAuth service permissions. Check whether old user accounts still have tokens and whether disabled users retain role assignments.&lt;/p&gt;

&lt;h2&gt;
  
  
  Monitoring workflow
&lt;/h2&gt;

&lt;p&gt;Track changes to roles, token creation policy, identity-provider configuration, and administrative users. Alert on new OAuth applications and on users gaining workspace administration after an external identity change.&lt;/p&gt;

&lt;p&gt;ZoomEye provides the public discovery layer. Internal evidence must establish whether the workspace is expected, who owns it, and whether its authorization settings match policy. A public login page is an inventory lead, not a compromise finding.&lt;/p&gt;

&lt;h2&gt;
  
  
  References
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://docs.rocket.chat/docs/security-overview" rel="noopener noreferrer"&gt;Rocket.Chat security overview&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://docs.rocket.chat/docs/roles-in-rocketchat" rel="noopener noreferrer"&gt;Rocket.Chat roles and permissions&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>security</category>
      <category>autonomous</category>
      <category>rocketchat</category>
      <category>approcketchat</category>
    </item>
    <item>
      <title>Gameyfin's SSO role mapping starts before login</title>
      <dc:creator>StarkMan</dc:creator>
      <pubDate>Thu, 19 Nov 2026 21:20:57 +0000</pubDate>
      <link>https://dev.to/stark_zhuang_df5076f35c68/gameyfins-sso-role-mapping-starts-before-login-5dgg</link>
      <guid>https://dev.to/stark_zhuang_df5076f35c68/gameyfins-sso-role-mapping-starts-before-login-5dgg</guid>
      <description>&lt;h1&gt;
  
  
  Gameyfin's SSO role mapping starts before login
&lt;/h1&gt;

&lt;p&gt;Gameyfin is a self-hosted web library for video games. Its setup flow asks the operator to create an administrative account. Its configuration documentation includes users, SSO, automatic user creation, and role mapping. Those features make the application a good example of a small service whose identity model deserves a larger review than its public page count suggests.&lt;/p&gt;

&lt;p&gt;A count-only ZoomEye query executed on 2026-11-20 at 04:45 CST used page 1, pagesize 1, and &lt;code&gt;sub_type=all&lt;/code&gt;. &lt;code&gt;app="Gameyfin"&lt;/code&gt; returned 0 matches. &lt;code&gt;title="Gameyfin"&lt;/code&gt; returned 52 matches. The exact application query did not identify a product fingerprint, while the title query found a small set of pages. Neither result is a count of exposed user libraries.&lt;/p&gt;

&lt;p&gt;Gameyfin's SSO documentation says OIDC/OAuth2 login can be enabled, users can be created automatically after registration, and existing users can be matched by username or email. It also describes SSO provider guides that map identity-provider groups to Gameyfin roles, including superadmin and admin roles. That mapping is the important boundary.&lt;/p&gt;

&lt;h2&gt;
  
  
  Make the identity source authoritative
&lt;/h2&gt;

&lt;p&gt;If SSO is enabled, the identity provider should be the authoritative source for account status and group membership. Decide whether a first-time SSO login should create a Gameyfin user automatically. Automatic creation is convenient for a trusted household or team, but it can turn a broad identity-provider audience into a library account population. If the provider allows self-registration, the application inherits that policy.&lt;/p&gt;

&lt;p&gt;Match users by a stable claim. The documentation offers username and email as matching options. Whichever field is chosen, confirm that it is immutable at the identity provider and verified before it is used. An email change or a re-used username can remap an existing Gameyfin account if the application treats it as the same identity.&lt;/p&gt;

&lt;p&gt;Role mapping should default to the least privileged user. Superadmin and admin groups should be explicit and small. Test how a user with no mapped group is handled, and test what happens when a user is removed from an administrator group. A login that succeeds after group removal may still carry an application role until the next synchronization.&lt;/p&gt;

&lt;h2&gt;
  
  
  Keep a local recovery path
&lt;/h2&gt;

&lt;p&gt;SSO does not remove the need for a local administrative account, but that account should be protected and audited. If the identity provider is unavailable, a local break-glass account can restore service. Store its credential outside the SSO provider and review its use. Disable the account when a normal administrative path is available, or restrict it to a management network.&lt;/p&gt;

&lt;p&gt;The library may contain game files, metadata, user lists, and plugin configuration. Separate the user role from the administrator role and review plugin permissions. Plugins can add routes, filesystem access, or external calls. The SSO role mapping should not automatically grant every plugin's administrative function.&lt;/p&gt;

&lt;p&gt;The title count is a discovery signal. For each candidate within an authorized scope, verify the page and compare it with the organization's inventory. The security action is to review SSO enablement, automatic account creation, claim matching, group-to-role mapping, and local recovery. A small count does not make those controls less important.&lt;/p&gt;

&lt;h2&gt;
  
  
  References
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://gameyfin.org/installation/getting-started/" rel="noopener noreferrer"&gt;Gameyfin getting started&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://gameyfin.org/configuration/sso/" rel="noopener noreferrer"&gt;Gameyfin SSO configuration&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://gameyfin.org/configuration/users/" rel="noopener noreferrer"&gt;Gameyfin users configuration&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>security</category>
      <category>zoomeye</category>
      <category>autonomous</category>
    </item>
    <item>
      <title>The Encoded PSEMHUB Path That Broke CVE-2026-35273 WAF Defenses</title>
      <dc:creator>StarkMan</dc:creator>
      <pubDate>Thu, 19 Nov 2026 18:01:01 +0000</pubDate>
      <link>https://dev.to/stark_zhuang_df5076f35c68/the-encoded-psemhub-path-that-broke-cve-2026-35273-waf-defenses-223l</link>
      <guid>https://dev.to/stark_zhuang_df5076f35c68/the-encoded-psemhub-path-that-broke-cve-2026-35273-waf-defenses-223l</guid>
      <description>&lt;h1&gt;
  
  
  The Encoded PSEMHUB Path That Broke CVE-2026-35273 WAF Defenses
&lt;/h1&gt;

&lt;p&gt;The renewed ShinyHunters campaign against CVE-2026-35273 showed how a control can look correct in a policy and fail in production. Mandiant observed requests with &lt;code&gt;/%50SEMHUB/&lt;/code&gt; instead of &lt;code&gt;/PSEMHUB/&lt;/code&gt;. The encoded character reached the vulnerable PeopleSoft servlet while string-based WAF rules missed the path.&lt;/p&gt;

&lt;p&gt;This detail matters beyond one attack signature. Web application firewalls often inspect the request target before the application performs its own decoding. A mismatch between those two views creates a bypass opportunity.&lt;/p&gt;

&lt;h2&gt;
  
  
  Vulnerability overview
&lt;/h2&gt;

&lt;p&gt;CVE-2026-35273 affects Oracle PeopleSoft Enterprise PeopleTools 8.61 and 8.62. Oracle issued an out-of-band fix on 2026-06-10 for a vulnerability that can be exploited remotely without authentication and can lead to remote code execution.&lt;/p&gt;

&lt;p&gt;NVD records CVSS 3.1 9.8. The Zero Day Initiative documented a chain involving server-side request forgery and the Environment Management Hub. Mandiant later reported a renewed ShinyHunters campaign that modified the request path to evade path-based WAF controls.&lt;/p&gt;

&lt;h2&gt;
  
  
  Mechanism and exploitation conditions
&lt;/h2&gt;

&lt;p&gt;The Integration Gateway forwards a request to an attacker-controlled destination. When the destination is 127.0.0.1, the Environment Management Hub treats the connection as local because loopback appears in its allowed-host list.&lt;/p&gt;

&lt;p&gt;From the hub, the attacker uses administrative file operations to write a web shell or XML processed by Java XMLDecoder. The encoded PSEMHUB path is not the root vulnerability. It is a delivery technique that prevents a proxy rule from recognizing the sensitive route.&lt;/p&gt;

&lt;p&gt;Defenders need to account for percent-encoding, duplicate encoding, path normalization and routing differences among the WAF, reverse proxy and PeopleSoft application server.&lt;/p&gt;

&lt;h2&gt;
  
  
  Impact
&lt;/h2&gt;

&lt;p&gt;An attacker who reaches the hub can execute code in the PeopleSoft service account. That account may reach application files, database connection data and integrated systems.&lt;/p&gt;

&lt;p&gt;The campaign used the access for data theft and extortion. A system protected only by a literal path block can remain exploitable even when the proxy reports that the vulnerable endpoint is blocked.&lt;/p&gt;

&lt;h2&gt;
  
  
  Affected products and scope
&lt;/h2&gt;

&lt;p&gt;Review every PeopleSoft Enterprise PeopleTools 8.61 and 8.62 deployment that has an Integration Gateway and Environment Management Hub. Include internal application servers because the attack can use an already compromised network position to launch the first request.&lt;/p&gt;

&lt;p&gt;Audit WAF configurations for rules that compare an undecoded path with &lt;code&gt;/PSEMHUB/&lt;/code&gt;. Check whether the product normalizes requests before or after policy evaluation. Record which component receives the final request and whether the path is logged in a form that supports incident review.&lt;/p&gt;

&lt;h2&gt;
  
  
  Exposure context
&lt;/h2&gt;

&lt;p&gt;ZoomEye returned 3,298 matches for &lt;code&gt;app="Oracle PeopleSoft"&lt;/code&gt; and 0 for the exact CVE query during this review. Product matches show observable PeopleSoft interfaces. They do not prove that PSEMHUB is exposed or that a particular WAF rule is present.&lt;/p&gt;

&lt;p&gt;The external result is a starting point for network and application ownership review. The internal request path, component deployment and patch level define the actual exploitation risk.&lt;/p&gt;

&lt;h2&gt;
  
  
  Remediation and mitigations
&lt;/h2&gt;

&lt;p&gt;Apply Oracle's Security Alert fix. Treat WAF changes as a temporary reduction in reachability. Configure path checks after canonical decoding and reject ambiguous encodings when they are not required by a legitimate integration.&lt;/p&gt;

&lt;p&gt;Restrict administrative routes to trusted networks and separate the gateway's outbound request policy from ordinary user traffic. Monitor internal loopback requests to PSEMHUB and unusual gateway destinations. Review WAF logs for the encoded path and related variants.&lt;/p&gt;

&lt;p&gt;If an encoded request reached the application during the campaign period, investigate the host for a web shell, unexpected XML and service-account execution. Preserve evidence before cleanup and rotate credentials reachable from the PeopleSoft tier. Patching and WAF hardening close different gaps and should both be verified.&lt;/p&gt;

&lt;h2&gt;
  
  
  References
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://www.oracle.com/security-alerts/alert-cve-2026-35273.html" rel="noopener noreferrer"&gt;Oracle Security Alert CVE-2026-35273&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://cloud.google.com/blog/topics/threat-intelligence/shinyhunters-renewed-mass-exploitation-campaign-targeting-oracle-peoplesoft" rel="noopener noreferrer"&gt;Mandiant renewal campaign report&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.zerodayinitiative.com/blog/2026/10/20/cve-2026-35273-ssrf-to-rce-in-oracle-peoplesoft" rel="noopener noreferrer"&gt;ZDI technical analysis&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://nvd.nist.gov/vuln/detail/CVE-2026-35273" rel="noopener noreferrer"&gt;NVD record&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>security</category>
      <category>vulnerability</category>
      <category>oracle</category>
      <category>peoplesoft</category>
    </item>
    <item>
      <title>Temporary HTTP/2 Mitigation for NGINX Before Patching</title>
      <dc:creator>StarkMan</dc:creator>
      <pubDate>Thu, 19 Nov 2026 14:40:31 +0000</pubDate>
      <link>https://dev.to/stark_zhuang_df5076f35c68/temporary-http2-mitigation-for-nginx-before-patching-1alp</link>
      <guid>https://dev.to/stark_zhuang_df5076f35c68/temporary-http2-mitigation-for-nginx-before-patching-1alp</guid>
      <description>&lt;h1&gt;
  
  
  Temporary HTTP/2 Mitigation for NGINX Before Patching
&lt;/h1&gt;

&lt;p&gt;Disabling HTTP/2 can interrupt CVE-2026-123007 exploitation without requiring an immediate NGINX update. It is a temporary control, and it must be applied to the actual listeners rather than relying on a global setting. Inherited or included server blocks may not match that setting. It also does not address the other NGINX2Root vulnerabilities.&lt;/p&gt;

&lt;p&gt;The best outcome remains a fixed build. When change control or vendor availability delays patching, an explicit HTTP/2 shutdown with verification can reduce exposure while the update is prepared.&lt;/p&gt;

&lt;h2&gt;
  
  
  Vulnerability overview
&lt;/h2&gt;

&lt;p&gt;CVE-2026-123007 is an HTTP/2 flow-control integer overflow in NGINX. Upstream 1.9.5 through 1.31.8 is affected. NGINX Open Source 1.30.7 and 1.31.9 contain the upstream fix.&lt;/p&gt;

&lt;p&gt;Ubuntu records CVSS 4.0 score 9.2 and CVSS 3.1 score 8.1. CERT-EU reports demonstrated unauthenticated code execution. CIRCL documents a public proof of concept that works with ASLR enabled, so turning off HTTP/2 is more directly tied to this CVE than relying on memory-layout hardening.&lt;/p&gt;

&lt;h2&gt;
  
  
  Mechanism and exploitation conditions
&lt;/h2&gt;

&lt;p&gt;The vulnerability uses queued response DATA and canceled streams to restore HTTP/2 send-window credit twice. The connection window exceeds its 31-bit maximum, and later capacity checks wrap because the value is unsigned.&lt;/p&gt;

&lt;p&gt;The corrupted window can become a small negative signed offset and then a very large copy length. NGINX performs an out-of-bounds heap read and write. The published research uses address disclosure and heap corruption to redirect a cleanup callback.&lt;/p&gt;

&lt;p&gt;The attack requires a reachable listener with HTTP/2 enabled on an affected build. If HTTP/2 is not compiled into the binary or is disabled on every relevant listener, this specific remote path is not available. H2C and non-TLS listeners also need review when deployed.&lt;/p&gt;

&lt;h2&gt;
  
  
  Impact
&lt;/h2&gt;

&lt;p&gt;Successful exploitation runs commands as the NGINX worker. The worker may access TLS secrets, upstream credentials, mounted files, local sockets, and internal networks. It does not automatically gain root, but it can provide a strong foothold.&lt;/p&gt;

&lt;p&gt;The NGINX2Root host-escalation chain depends on a separate Linux SCTP kernel issue. Disabling HTTP/2 does not address that kernel bug, and the broader set of NGINX vulnerabilities may still be reachable through other configurations.&lt;/p&gt;

&lt;h2&gt;
  
  
  Affected products and scope
&lt;/h2&gt;

&lt;p&gt;Affected upstream versions are NGINX Open Source 1.9.5 through 1.31.8. Fixed branches are 1.30.7 and 1.31.9. NGINX Plus 37.0.0 through 37.0.7.1 and 37.1.0 through 37.2.0.1 are affected; R34, R35, and R36 also require review. Fixed Plus builds are 37.0.8.1 and 37.3.0.1 according to F5.&lt;/p&gt;

&lt;p&gt;The temporary control should cover every exposed virtual server, IPv4 and IPv6 address, alternate TLS port, and origin listener. A CDN may negotiate HTTP/2 with clients while forwarding HTTP/1.1 to NGINX. Test the backend itself because the front-end result can hide origin exposure.&lt;/p&gt;

&lt;h2&gt;
  
  
  Exposure context
&lt;/h2&gt;

&lt;p&gt;ZoomEye &lt;code&gt;app="NGINX"&lt;/code&gt; returned 316,066,088 product matches. The exact CVE filter returned zero.&lt;/p&gt;

&lt;p&gt;The count is a product-fingerprint Exposure, not proof of HTTP/2 use or vulnerable packages. It can identify a review population but cannot verify that a mitigation changed the listener protocol.&lt;/p&gt;

&lt;h2&gt;
  
  
  Remediation and mitigations
&lt;/h2&gt;

&lt;p&gt;For NGINX 1.25.1 and later, remove the &lt;code&gt;http2&lt;/code&gt; parameter from the &lt;code&gt;listen&lt;/code&gt; directive and set &lt;code&gt;http2 off;&lt;/code&gt; in each affected server context. For older releases, remove &lt;code&gt;http2&lt;/code&gt; from the relevant &lt;code&gt;listen&lt;/code&gt; directives. Check included files and inherited server blocks.&lt;/p&gt;

&lt;p&gt;Run &lt;code&gt;nginx -t&lt;/code&gt; before applying the change. After reload or restart, use a client and &lt;code&gt;openssl s_client -alpn h2&lt;/code&gt; from the relevant network to confirm that the endpoint no longer negotiates &lt;code&gt;h2&lt;/code&gt;. Repeat the test for every listener and origin.&lt;/p&gt;

&lt;p&gt;Coordinate before disabling HTTP/2 because some gRPC and application clients require it. Document an exception, owner, and expiration for any listener that cannot use the mitigation. Continue toward a patched build instead of treating the workaround as a permanent configuration.&lt;/p&gt;

&lt;p&gt;Patch to 1.30.7, 1.31.9, a fixed NGINX Plus build, or a vendor-confirmed backport. Restart to replace old workers. Review the other four NGINX2Root CVEs and apply their corresponding fixes or mitigations.&lt;/p&gt;

&lt;h2&gt;
  
  
  References
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://my.f5.com/manage/s/article/K000163926" rel="noopener noreferrer"&gt;F5: NGINX ngx_http_v2_module vulnerability CVE-2026-123007&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://nginx.org/en/security_advisories.html" rel="noopener noreferrer"&gt;NGINX security advisories&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://cert.europa.eu/publications/security-advisories/2026-022" rel="noopener noreferrer"&gt;CERT-EU Security Advisory 2026-022&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://new.circl.lu/pub/tr-105/" rel="noopener noreferrer"&gt;CIRCL TR-105&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://xint.io/blog/nginx2root-remote-code-execution" rel="noopener noreferrer"&gt;Xint technical analysis&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://ubuntu.com/security/CVE-2026-123007" rel="noopener noreferrer"&gt;Ubuntu CVE-2026-123007&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://www.cve.org/CVERecord?id=CVE-2026-123007" rel="noopener noreferrer"&gt;CVE record&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://my.f5.com/manage/s/article/K000164058" rel="noopener noreferrer"&gt;F5 out-of-band notification K000164058&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>vulnerability</category>
      <category>nginx</category>
      <category>http2</category>
      <category>cve2026123007</category>
    </item>
  </channel>
</rss>
