HispasecAugust 25, 2026🇪🇸Translated from Spanish

Keycloak Fixes Critical CVE-2026-18963 Password Reset Flaw Allowing Unauthenticated Account Takeover

Keycloak has corrected a critical vulnerability, CVE-2026-18963, that allows a remote unauthenticated attacker to force a password reset for any user and take over the account. The priority is to update to the fixed versions or, if that is not possible, disable the Forgot password feature in all realms.

Keycloak, one of the most widely deployed products in corporate environments for identity and access management (IAM), has closed a critical vulnerability that opens the door to account takeover without authentication. The flaw, registered as CVE-2026-18963 and carrying a CVSS 3.1 score of 9.1, affects the reset-credentials flow and permits a remote attacker to complete a user password change without the usual email verification step.

The weakness stems from incorrect state validation inside the authentication flow. In practical terms, the system may accept a sequence of steps that should remain blocked until the user confirms the link sent by email. The result is a bypass: the attacker skips the verification phase and reaches the password change screen or endpoint directly, with the ability to set a new password and later access the account as the legitimate owner.

The most sensitive scenario occurs when the attacker targets administrative accounts. With that control, the attacker can modify policies, create users, assign roles, or alter configurations that affect applications integrated with Keycloak. The CVSS metric accurately reflects the risk: network attack vector, low complexity, no prior privileges, and no user interaction required. The impact focuses on confidentiality and integrity, while availability is not directly affected.

Upstream patches have been released in Keycloak 26.7.2 dated 19 August 2026. For Red Hat Build of Keycloak (RHBK), fixes are provided through updates in the 26.4 and 26.6 branches, specifically versions 26.4.15 and 26.6.6. As of 24 August 2026, Red Hat reports no known exploitation in the wild and no verified public exploit.

For organizations unable to apply updates immediately, the clearest mitigation is to disable the Forgot password recovery option. This setting must be applied realm by realm, so administrators should verify that the feature is disabled in every realm, especially in environments with multiple realms for subsidiaries, customers, or internal applications.

After applying the patch or mitigation, auditing remains essential. Reviewing login events and credential change logs for privileged accounts helps detect anomalous password resets and limits the scope if an attacker attempted to exploit the exposure window. In IAM systems, such a flaw can affect the entire connected ecosystem.

Related articles

HabrVulnerabilities & Exploits

Critical MikroTik RouterOS Vulnerabilities Enable SSH Authentication Bypass and Privilege Escalation

Polish CERT disclosed three vulnerabilities in MikroTik routers, two of which have been actively exploited since at least September 2. The flaws, rated 9.2 on CVSS, affect devices with internet-facing SSH access and were discovered using OpenAI models GPT 5.5 Cyber and GPT 5.6 Sol followed by manual verification. CVE-2026-67276 allows authentication bypass when the attacker knows the username and public key module, while CVE-2026-86060 permits privilege escalation via usernames containing invalid characters. Their combination enables full device compromise. A third issue, CVE-2026-67277 rated 8.8, resides in the speed-test service and can cause denial of service. Patches are available in RouterOS versions 7.25beta3, 7.24.2, 7.23.4 and 6.49.21, and MikroTik added detection for prior compromise. Attacks began concurrently with patch release and leave distinctive log entries.

HabrVulnerabilities & Exploits

Yookassa Webhook Flaw Allowed Forging Payment Confirmations via Single Curl Request

A technical audit of a car rental service uncovered a critical flaw in the Yookassa webhook handler where payment.succeeded events were trusted without verification. The endpoint accepted the event type and gateway_payment_id directly from the client-supplied payload, enabling attackers to mark pending bookings as paid using a simple unauthenticated POST request. The vulnerability existed because the integration did not use Yookassa's optional Webhook-Signature mechanism and avoided IP allowlisting due to proxy infrastructure concerns. Developers addressed the issue by implementing a reverse API call to fetch authoritative payment status from Yookassa before updating records, ensuring fail-closed behavior on errors. New tests cover forgery attempts, amount mismatches, and idempotency, while several additional hardening measures were deliberately deferred to keep the patch minimal and reviewable. The original unit tests had masked the problem by simulating trusted input rather than validating source authenticity.

SecuritylabVulnerabilities & Exploits

Password Deleted from Git but Still Present: Major CI/CD Security Mistakes

Developers often leave secrets in configuration files, .env examples, or test scripts that end up in Git history. Removing a file in a new commit does not erase previous versions, and .gitignore offers no protection for already tracked content. Pre-commit hooks with tools like Gitleaks can scan staged changes before they reach the repository. Docker images can retain secrets in layers even after removal commands, requiring BuildKit secret mounts instead. Terraform plans and state files may still expose sensitive values marked only as sensitive. Kubernetes workloads need both non-root containers with dropped capabilities and narrowly scoped RBAC roles. Multiple scanning stages from pre-commit through post-deployment help catch issues before they reach production.

Security NEXTVulnerabilities & Exploits

Critical SSRF Vulnerability in AWS SSM Agent Allows IAM Credential Theft via Port Forwarding Bypass

Amazon Web Services has disclosed a high-severity server-side request forgery vulnerability in the AWS Systems Manager Agent. The flaw, tracked as CVE-2026-89049, affects the port forwarding feature used by Session Manager and stems from insufficient validation that permits bypass of deny-list restrictions on link-local addresses. Successful exploitation requires port-forwarding permissions but can lead to unauthorized access to instance metadata and temporary IAM role credentials. The company rated the issue as Important with CVSSv4.0 base score 8.5 and CVSSv3.1 score 9.9. The vulnerability was addressed in version 3.3.4851.0 released on 13 July 2026, prior to the public advisory issued on 10 September 2026.