CVE-2026-20211 is a critical vulnerability in Cisco Identity Services Engine (ISE) that allows an authenticated attacker with high-privileged administrative credentials to execute arbitrary commands on the underlying operating system. The flaw arises from insecure deserialization of Java objects. Exploitation involves sending a crafted serialized Java object to the affected device, potentially leading to user-level access and privilege escalation to root. In single-node deployments, exploitation can cause a denial-of-service condition by making the node unavailable, disrupting network access for unauthenticated endpoints.
On affected platforms running Arista EOS with OpenConfig-related services (i.e., gNMI, gNSI, RESTCONF and NETCONF), sensitive requests and responses may be unintentionally logged. These may be stored on the local EOS device or recorded on remote accounting servers. Note that gRPC-based streaming via Streaming Telemetry Agent to CloudVision is not affected by this vulnerability.
Examples of sensitive information include:
- Sensitive CLI commands (e.g., "username bob secret myPass")
- Sensitive OpenConfig YANG leafs (e.g., "system/aaa/global/tacacs/config/secret-key")
This issue was discovered internally by Arista, and the company is not aware of any malicious exploitation of this vulnerability in customer networks.
CVE-2026-58196 is a Server-Side Request Forgery (SSRF) vulnerability in stacklok's toolhive utility versions prior to 0.31.0. The flaw arises from the handling of authentication discovery URLs, where the server-side HTTP clients trust remote-server-controlled destinations and follow redirects without proper restrictions. This allows a malicious or compromised MCP server to induce the toolhive client to make unauthorized requests to internal-only HTTP services, including link-local and RFC1918 addresses. The vulnerability can be exploited to access internal metadata services such as AWS IMDSv1, potentially leaking sensitive credentials. The issue is fixed in version 0.31.0.
Check Point Research discovered a covert cross-account communication channel in ChatGPT's code-execution sandbox environment. This channel allows an attacker to execute hidden tasks within a victim's ChatGPT session, leveraging the victim's connected tools and data without revealing the attack in the visible conversation. The vulnerability arises from a shared internal service used for software package delivery that unintentionally enables communication between isolated containers of different accounts. The attack can exfiltrate sensitive data such as emails, conversation history, and files accessible in the victim's session. The scope depends on the victim's session permissions and connected services. This issue highlights an architectural weakness in isolation boundaries within ChatGPT's sandboxed environments.
Forgejo versions 15.0.0 through 15.0.4 contain a server-side request forgery (SSRF) vulnerability in the Repository Migration Handler component. The issue arises from the net.LookupIP function in the file services/migrations/allowlist/is_migrate_allowed.go. This vulnerability can be exploited remotely without user interaction. A patch exists but is a breaking change and will not be backported to versions 15 or 16.
CVE-2026-55673 is an OS command injection vulnerability in powsybl-core prior to version 7.2.2. It arises from insufficient escaping of command arguments and environment variables in UnixLocalCommandExecutor and WindowsLocalCommandExecutor, allowing attacker-controlled input to execute arbitrary shell commands as the JVM user. This affects multiple components including action-simulator, security-analysis, and dynamic-security-analysis, and can be exploited remotely via downstream tools and services that forward untrusted parameters. The issue is fixed in version 7.2.2.
CVE-2026-77573 is a time-of-check to time-of-use (TOCTOU) race condition vulnerability in Weblate versions prior to 2026.8. It allows a user with permission to manage component repository URLs to perform server-side request forgery (SSRF) against internal services via DNS rebinding during version control system (VCS) operations. The vulnerability arises because Weblate validates the hostname's initial DNS resolution but the external VCS clients perform a separate DNS lookup later, enabling an attacker-controlled hostname to be redirected to internal or private addresses. This can lead to unauthorized access to internal VCS-compatible services and potential exposure of private repository contents. The issue affects installations that allow untrusted repository hostnames while using the VCS_RESTRICT_PRIVATE=True setting. The vulnerability is fixed in Weblate version 2026.8.
CVE-2026-66908 is an improper authentication vulnerability in the Apache Camel Platform HTTP Main component affecting versions from 4.8.0 up to but not including 4.22.0. The issue arises because the JWT authentication configuration does not enforce validation of the issuer (iss) and audience (aud) claims unless explicitly configured, allowing acceptance of any unexpired token signed by a trusted key. This can lead to acceptance of tokens intended for other services or audiences within the same trust domain. The vulnerability is fixed in version 4.22.0, which requires explicit configuration of issuer and audience claims or an explicit override to allow missing claims. Users on maintenance branches 4.14.x and 4.18.x should upgrade to versions 4.14.9 or 4.18.4 and configure these claims accordingly.
CVE-2026-77775 is a Server-Side Request Forgery (SSRF) vulnerability in Headroom Labs' Headroom product. The vulnerability arises because the LLM proxy component allows clients to specify the upstream destination via the x-headroom-base-url header without proper validation. This enables attackers to make requests to internal services, including cloud metadata endpoints, and receive their responses. The default deployment configuration exposes these routes without authentication, increasing the risk of exploitation. The issue is fixed in version 0.36.1.
CVE-2026-72818 is a high-severity vulnerability in the nltk library's TweetTokenizer component. It arises from an inefficient regular expression in nltk/tokenize/casual.py that can cause exponential backtracking when processing certain crafted inputs. This leads to excessive CPU consumption and potential denial of service by stalling services that tokenize untrusted social media text. Version 3.10.1 of nltk addresses this issue by bounding the label repetition in the regular expression.
CVE-2026-76392 is a medium severity vulnerability in Splunk AI Toolkit versions 5.7 and all versions from 5.7 up to but not including 6.0.0. It involves the use of hard-coded or predictable credentials for connected container services, which can be accessed by users without admin or power roles. This issue arises because the toolkit generates or stores credentials using default or predictable values, potentially allowing unauthorized access to these services.
ArcadeDB versions before 26.8.1 have a server-side request forgery (SSRF) vulnerability in the IMPORT DATABASE command. The vulnerability arises because the security validator resolves and checks hostnames, but the actual connection re-resolves the raw URL and follows redirects. Authenticated attackers can exploit this to bypass validation using DNS rebinding or HTTP redirects, potentially accessing cloud metadata endpoints, internal services, or reading arbitrary local files on default installations.
OpenClaw CLI contains a local command injection vulnerability in its systemd unit file generation on Linux. The vulnerability arises because attacker-controlled environment variable values are not properly sanitized for newline characters, allowing injection of additional systemd directives. When the service is restarted, these injected directives execute arbitrary commands with the privileges of the OpenClaw gateway service user. This affects Linux deployments using systemd user services where an attacker can influence the environment variables and trigger service reinstall or install. A fix for this vulnerability is available.
mcp-server-kubernetes versions 2.9.1 up to but not including 3.5.0 contain an argument injection vulnerability in the port_forward tool. This flaw arises from unsafe string concatenation and naive space splitting of user input when constructing kubectl commands, allowing injection of arbitrary kubectl flags. Exploitation can expose internal Kubernetes services to the network, enable cross-namespace targeting, and facilitate indirect attacks via prompt injection against AI agents connected to the MCP server. The vulnerability is fixed in version 3.5.0.
CVE-2025-13394 is a Cross-Site Request Forgery (CSRF) vulnerability in the Ajax processor of the Carbon console within WSO2 Identity Server. The vulnerability arises because state-changing operations use the HTTP GET method and rely on the SameSite=Lax cookie attribute for CSRF mitigation, which can be bypassed. This allows attackers to trick authenticated users into unknowingly performing unauthorized actions. Exploitation requires the Carbon console and related services to be publicly accessible, which is against WSO2's security recommendations.
A flaw was found in Keycloak's Authorization Services. The component responsible for matching request paths to security policies (PathMatcher) does not properly normalize URIs before comparison. By adding extra characters like a trailing slash or matrix parameters to a URL, an attacker can trick the system into applying a less restrictive security policy than intended. This allows an authenticated user to access administrative or restricted areas they should not have permission to see.
The VikAppointments Services Booking Calendar WordPress plugin contains an unauthenticated SQL injection vulnerability. This flaw arises because a parameter controlling the sorting of the public reviews list is used directly in a database query without proper validation or sanitization. An attacker can exploit this to inject arbitrary SQL commands and read sensitive data such as WordPress user credentials without needing to authenticate or have special privileges.
A VCO feature does not sufficiently validate caller-supplied input, allowing requests to be made on behalf of authenticated tenant accounts to internal services that are not otherwise accessible. This vulnerability requires a minimum role of Enterprise Standard Admin.
This issue was discovered internally by Arista and the company is not aware of any malicious uses of this issue in customer networks.
FFmpeg versions through 8.1.2 contain an information disclosure vulnerability in the LCL/ZLIB video decoder. This flaw allows attackers to expose uninitialized heap memory by providing a valid zlib stream that decompresses to fewer bytes than the expected frame size. The vulnerability arises because the decompression function treats short decompression as non-fatal and copies a full frame's worth of data from the allocation buffer, potentially leaking sensitive memory contents. This can aid attackers in bypassing address space layout randomization (ASLR) in long-running media processing services. A fix is available in commit 8670835.