CVE Feed

Envoy is an open source edge and service proxy designed for cloud-native applications. From 1.37.0 until 1.37.5 and 1.38.3, when the %REQUESTED_SERVER_NAME(X:Y)% is used in log format and host related options is specified, like HOST_FIRST, SNI_FIRST, it's possible to crash Envoy when the specified host header is missing in the request headers. This vulnerability is fixed in 1.37.5 and 1.38.3.

Envoy is an open source edge and service proxy designed for cloud-native applications. From 1.36.0 until 1.36.9, 1.37.5, and 1.38.3, a Use-After-Free (UAF) vulnerability leading to a sudden segmentation fault exists in Envoy's ext_authz HTTP filter when processing per-route authorization overrides concurrently with rapid downstream client disconnects. During standard request lifecycles, Envoy instantiates the ext_authz filter with a foundational authorization client object (client_). If a matched route dictates a dynamic per-route HTTP or gRPC authorization service override, the filter generates a localized client. In the vulnerable implementation, this transient client aggressively overwrote the default client_ unique pointer by executing client_ = std::move(per_route_client). When a client rapidly establishes and subsequently tears down a stream (such as rapidly refreshing a protected WebSocket endpoint), the downstream triggers the ConnectionManagerImpl::doDeferredStreamDestroy() -> ActiveStream::onResetStream() lifecycle. Envoy immediately sequences Filter::onDestroy() in an attempt to securely abort dispatched asynchronous authorization check transactions via client_->cancel(). By destructing the default client abruptly during initiateCall, a memory lifecycle misalignment occurs within the async client manager. The stream teardown fails to reliably track and cancel the dynamically bound asynchronous authorization tasks, orchestrating a sequence where a late asynchronous callback from the network evaluates against a heavily destroyed ActiveStream validation span, generating a UAF process crash. This vulnerability is fixed in 1.36.9, 1.37.5, and 1.38.3.

Envoy is an open source edge and service proxy designed for cloud-native applications. From 1.34.0 until 1.35.13, 1.36.9, 1.37.5, and 1.38.3, PROXY Protocol v2 header generator emits TLVs beyond the maximum length of 65535 bytes, causing a mismatch between bytes written and the length field in the header. This can result in smuggled bytes on the upstream request. This vulnerability is fixed in 1.35.13, 1.36.9, 1.37.5, and 1.38.3.

Envoy is an open source edge and service proxy designed for cloud-native applications. From 1.34.0 until 1.35.13, 1.36.9, 1.37.5, and 1.38.3, Envoy crashes if an ext_proc server sends a single gRPC message containing multiple, specially crafted ProcessingResponse messages. This can occur when the first response in the batch causes the gRPC stream object to be destroyed, leading to a use-after-free error when Envoy attempts to process subsequent responses in the same gRPC message. This vulnerability is fixed in 1.35.13, 1.36.9, 1.37.5, and 1.38.3.

Envoy is an open source edge and service proxy designed for cloud-native applications. From 1.34.0 until 1.35.13, 1.36.9, 1.37.5, and 1.38.3, a vulnerability exists in Envoy's TCP StatsD sink (TcpStatsdSink), where the thread-local flusher buffer can be overflowed by exceptionally long statistic names (e.g., >16KiB). During formatting, TcpStatsdSink reserves a single contiguous memory slice of 16KiB (FLUSH_SLICE_SIZE_BYTES). If formatting a single metric exceeds the remaining capacity, the flusher initiates a buffer rotation but incorrectly continues to allocate another fixed 16KiB slice. If an attacker can trigger a statistic name longer than 16KiB—for example, by sending an HTTP or gRPC request with an extremely long request path (:path) that is recorded by the grpc_stats filter configured with stats_for_all_methods: true—the flusher will attempt to copy the metric name using memcpy operations beyond the allocated heap buffer boundaries. This leads to a heap write overflow, which can cause immediate denial-of-service (process crash) or potential remote code execution (RCE). This vulnerability is fixed in 1.35.13, 1.36.9, 1.37.5, and 1.38.3.

Envoy is an open source edge and service proxy designed for cloud-native applications. From 1.26.0 until 1.35.13, 1.36.9, 1.37.5, and 1.38.3, the envoy.filters.http.grpc_stats filter crashes (null pointer dereference / segfault) when a Connect protocol request (Content-Type: application/connect+proto or application/connect+json) hits a direct_response route. A single unauthenticated HTTP request crashes the Envoy process. This vulnerability is fixed in 1.35.13, 1.36.9, 1.37.5, and 1.38.3.

Envoy is an open source edge and service proxy designed for cloud-native applications. From 1.18.0 until 1.35.13, 1.36.9, 1.37.5, and 1.38.3, the router filter contains a null pointer dereference vulnerability when handling HTTP 303 (See Other) internal redirects for body-less non-GET/HEAD requests. When a POST, PUT, DELETE, or PATCH request without a body is sent to a route configured with internal redirect policy that includes 303 in redirect_response_codes, and the upstream responds with HTTP 303, the redirect handling code attempts to drain a request body buffer that was never allocated. This results in a segmentation fault that crashes the entire Envoy process. When route configured with internal_redirect_policy including 303 in redirect_response_codes and upstream must return HTTP 303 response, an unauthenticated attacker can exploit this to cause complete denial of service, terminating all active connections. This vulnerability is fixed in 1.35.13, 1.36.9, 1.37.5, and 1.38.3.

Envoy is an open source edge and service proxy designed for cloud-native applications. Prior to 1.35.11, 1.36.7, 1.37.3, and 1.38.1, Envoy can translate a downstream HTTP/3 request that is complete at the transport layer (HEADERS with FIN / headers-only close) but still carries a nonzero Content-Length into a complete upstream HTTP/1 request with unresolved body debt. In an HTTP/1 upstream deployment where the origin replies before reading the declared body and keeps the connection reusable, the beginning of the next Envoy-generated upstream request can be consumed as the first request's body. The remaining bytes are then parsed by the origin as a new HTTP/1 request. This was reproduced as a route-bypass/desync: direct /pwn was denied by Envoy, but the second downstream H3 stream received the response for backend-parsed GET /pwn HTTP/1.1. This vulnerability is fixed in 1.35.11, 1.36.7, 1.37.3, and 1.38.1.

Envoy is an open source edge and service proxy designed for cloud-native applications. Prior to 1.35.11, 1.36.7, 1.37.3, and 1.38.1, in cases where UDP DNS filter is configured with local resolution containing a name with the length of 255 octets or remote resolution for a name of 255 octets long can complete successfully, a query with such name will result in abnormal process termination. The abnormal process termination is triggered by an invalid runtime precondition that the query name is strictly less than 255 octets, contradicting DNS specification rfc1035#section-2.3.4 that the name can be 255 or less octets. This vulnerability is fixed in 1.35.11, 1.36.7, 1.37.3, and 1.38.1.

Envoy is an open source edge and service proxy designed for cloud-native applications. From 1.23.0 until 1.35.11, 1.36.7, 1.37.3, and 1.38.1, a vulnerability has been identified in Envoy's zstd decompressor implementation (ZstdDecompressorImpl). When zstd decompression is enabled, processing a specially crafted, highly compressed zstd payload can lead to massive memory allocation. An attacker can exploit this to cause severe memory exhaustion, potentially resulting in an Out-Of-Memory (OOM) kill and Denial of Service (DoS) for the Envoy proxy. This vulnerability is fixed in 1.35.11, 1.36.7, 1.37.3, and 1.38.1.

Envoy is an open source edge and service proxy designed for cloud-native applications. Prior to 1.35.11, 1.36.7, 1.37.3, and 1.38.1, destructor of JSON Object results in stack overflow when deeply O(100K) nested objects are present. This vulnerability is fixed in 1.35.11, 1.36.7, 1.37.3, and 1.38.1.

Envoy is an open source edge and service proxy designed for cloud-native applications. Prior to 1.35.11, 1.36.7, 1.37.3, and 1.38.1, a structural flaw was identified in DefaultCertValidator::verifySubjectAltName where the extracted DNS SAN string is cast to a C-style string using .c_str() before being passed to the Utility::dnsNameMatch() algorithm. If the attacker serves a certificate with a dNSName SAN containing an embedded NUL byte, the helper Utility::generalNameAsString captures the complete string including the NUL. However, when .c_str() evaluates it, implicit conversion to absl::string_view inside dnsNameMatch relies on strlen(), prematurely truncating the evaluation context. Envoy evaluates trucated string against the exact required config_san match and returns true, thereby successfully validating the string with the Nul byte for an upstream routing. This vulnerability is fixed in 1.35.11, 1.36.7, 1.37.3, and 1.38.1.

Envoy is an open source edge and service proxy designed for cloud-native applications. Prior to 1.35.11, 1.36.7, 1.37.3, and 1.38.1, the OAuth2 HTTP filter's encrypt()/decrypt() functions use AES-256-CBC without an authentication tag (no HMAC, no AEAD). The /callback endpoint returns HTTP 302 on successful decryption and HTTP 401 on padding failure, creating a padding oracle. An attacker who obtains the encrypted CodeVerifier cookie can recover the plaintext PKCE code_verifier in ~6,200 requests (~100 seconds), then exchange it with a stolen authorization code to obtain the victim's access token. This vulnerability is fixed in 1.35.11, 1.36.7, 1.37.3, and 1.38.1.

mise manages dev tools like node, python, cmake, and terraform. Prior to 2026.3.10, mise processes .tool-versions files through the Tera template engine during parsing, with the exec() function registered, enabling arbitrary command execution. Unlike .mise.toml files, .tool-versions files are not subject to trust verification in non-paranoid mode. This means an attacker can place a malicious .tool-versions file in a git repository, and when a victim with mise activated cds into the directory, arbitrary commands execute without any trust prompt. This vulnerability is fixed in 2026.3.10.

mise manages dev tools like node, python, cmake, and terraform. Prior to 2026.6.4, mise's trust feature gates config files (mise.toml, .tool-versions) through trust_check, but task-include files are loaded on a path that never reaches it. When a directory has a task-include dir (mise-tasks/, .mise/tasks/, …) but no config file, mise falls back to the default includes and renders each task's tera fields — and that tera environment has exec() registered. A {{ exec(command='…') }} in any rendered field runs arbitrary commands the moment the tasks are merely listed. There's no config file to gate on, so no trust prompt ever appears. Read-only commands trigger it: mise tasks, mise task ls, mise run, mise tasks --usage (the query shell completion runs on Tab). The victim only has to cd into a cloned repo and list or tab-complete a task. This vulnerability is fixed in 2026.6.4.

mise manages dev tools like node, python, cmake, and terraform. Prior to 2026.6.1, the mise HTTP backend builds its install symlink destination from the raw resolved version string for non-latest versions. Normal tool install paths use the sanitized version pathname, but the HTTP backend's symlink path uses the raw value. On Unix-like systems, if that version is an absolute path, PathBuf::join discards the intended mise installs root. A repository-controlled .tool-versions file can therefore make mise install create a symlink outside the mise install tree. With bin_path, the same issue can place an executable symlink under an attacker-selected absolute prefix, such as a developer-tool prefix that is later added to PATH. This vulnerability is fixed in 2026.6.1.

mise manages dev tools like node, python, cmake, and terraform. From 2026.3.15 until 2026.6.4, mise loads github.credential_command from local project config before any trust decision, then executes that value with sh -c when resolving a GitHub token. An attacker who can place a .mise.toml in a repository can execute arbitrary shell commands when the victim runs a GitHub-related mise command and no higher-priority GitHub token environment variable is set. This vulnerability is fixed in 2026.6.4.

extract-zip does not validate symlink targets when extracting zip archives. When processing a malicious zip file containing a symlink with a relative path like '../../../../etc/passwd', extract-zip will extract the symlink without validation, allowing it to point outside the extraction directory. Depending on how extract-zip is used, an attacker could read or write to arbitrary files.

Dragonfly is an in-memory data store built for modern application workloads. Prior to 1.39.0, a crafted RESTORE payload triggers an out-of-bounds read in DragonflyDB's listpack collection loaders, crashing the entire server process (SIGSEGV). Because DragonflyDB requires no authentication by default and RESTORE is a normal keyspace command, an unauthenticated remote attacker can crash the server with a single ~24-byte command — a remote, repeatable denial of service. This vulnerability is fixed in 1.39.0.

Dragonfly is an in-memory data store built for modern application workloads. Prior to 1.39.9, Dragonfly has a RESP Protocol Injection via Lua redis.error_reply() in EvalSerializer. An authenticated user can inject arbitrary RESP messages into the connection's response stream, potentially causing response desynchronization in connection-pool clients. This vulnerability is fixed in 1.39.9.

GitHub MCP Server is GitHub's official MCP Server. From 0.22.0 until 1.1.2, when running in HTTP mode with --lockdown-mode enabled, the RepoAccessCache is implemented as a process-global singleton initialized with the first authenticated user's GraphQL client. All subsequent requests from different users share this singleton and their lockdown-related GraphQL queries are executed using the first user's credentials. The singleton is never updated to reflect later users' tokens. This vulnerability is fixed in 1.1.2.

Podman is a tool for managing OCI containers and pods. From 3.0.0 until 5.7.1, running a malicious container image where the WORKDIR path contains a symlink can create a directory or modify ownership on the host filesystem. Modified ownership is less likely to happen as that requires help from an untrusted/malicious process that mutates the host filesystem tree during dereferencing of the WORKDIR path, to trigger a race condition. This vulnerability is fixed in 5.7.1.

Podman is a tool for managing OCI containers and pods. From 1.8.1 until 5.8.4, a container image that contains a environment variable with just a key and no value can trick podman into passing that variable from the host into the container. This is made worse by the fact that using an asterisk (*) will cause podman to pass all host variables into the container. So essentially a malicious image can exfiltrate all podman environment variables that are set in the session from where the container is launched. This vulnerability is fixed in 5.8.4 and 6.0.0.

Dokku is a docker-powered PaaS. Prior to 0.38.7, the cron plugin utilizes commands in the app.json file to manage system cron running as the Dokku user. An app.json cron command utilizing special shell characters - including, but not limited to, > or ; - can break out of the Docker container and execute commands on the host as the Dokku user. This vulnerability is fixed in 0.38.7.

In Canonical LXD versions 4.12 through 6.9, a Server-Side Request Forgery (SSRF) vulnerability in the image import functionality allows authenticated users with the can_create_images entitlement to interact with internal network infrastructure via the /images endpoint. When importing an image from a URL source, the LXD daemon fails to validate or restrict outbound destination IP addresses, allowing connections to loopback, RFC1918 private ranges, and cloud metadata endpoints. This enables error-based port scanning and unauthorized interaction with internal HTTP services from the daemon's network position.

Dokku is a docker-powered PaaS. Prior to 0.38.2, the git:from-archive and certs:add commands extract user-supplied tar/zip archives into temporary directories without sanitizing member paths or preventing symlink traversal. GNU tar creates symlinks during extraction and follows them for subsequent entries, allowing an attacker to write arbitrary files anywhere writable by the dokku user — including overwriting ~/.ssh/authorized_keys to gain unrestricted shell access. This vulnerability is fixed in 0.38.2.

Dokku is a docker-powered PaaS. Prior to 0.38.2, the openresty-vhosts plugin copies files from an app's openresty/http-includes/ git repository directory to the host and then interpolates their filenames, unescaped, into a single-quoted shell string that is later parsed by eval. A filename containing a single quote breaks the quoting and allows command substitution to execute arbitrary commands on the host as the dokku user during the app's next deploy. This vulnerability is fixed in 0.38.2.

Dokku is a docker-powered PaaS. Prior to 0.38.2, the git:auth command creates $DOKKU_ROOT/.netrc using bash's touch command, which applies the default umask of 0644. This pre-creation defeats the netrc binary's built-in 0600 permission setting, leaving git credentials readable by any local user who can traverse the dokku home directory. This vulnerability is fixed in 0.38.2.

Dokku is a docker-powered PaaS. Prior to 0.38.2, the app name validation regex (^[a-z0-9][^/:_A-Z]*$) permits shell metacharacters. When an authenticated user pushes to a git remote with a crafted app name, the name is embedded unquoted into a bash pre-receive hook script via an unquoted heredoc (<<EOF instead of <<'EOF') in fn-git-create-hook() at plugins/git/internal-functions:378. On git push, bash interprets the semicolon as a command separator, executing arbitrary commands as the dokku user. This vulnerability is fixed in 0.38.2.

Echo is a Go web framework. Prior to 4.15.3 and 5.2.0, Echo's router and static file handler disagree on URL path decoding. The router matches routes using the raw encoded path (preserving %2F as-is), while StaticDirectoryHandler unescapes %2F to / before resolving filesystem paths. This allows an attacker to bypass route-level access controls and read static files without authorization. This vulnerability is fixed in 4.15.3 and 5.2.0.

AutoGPT is a workflow automation platform for creating, deploying, and managing continuous artificial intelligence agents. Prior to 0.6.32, there is a DoS vulnerability in AITextSummarizerBlock. Malicious users can amplify their input. For example, if a malicious user inputs 10K of content, the server will consume 50G of memory, eventually causing memory resources to be exhausted, resulting in DoS. This vulnerability is fixed in 0.6.32.

An Improper Authorization vulnerability exists in PayloadCMS version 3.84.1 due to insufficient access control on the account unlock operation.

AutoGPT is a workflow automation platform for creating, deploying, and managing continuous artificial intelligence agents. Prior to 0.6.32, there is a DoS vulnerability in ExtractTextInformationBlock. Malicious users can amplify their input. For example, if a malicious user inputs 10K of content, the server will consume 50G of memory, eventually causing memory resources to be exhausted, resulting in DoS. This vulnerability is fixed in 0.6.32.

AutoGPT is a workflow automation platform for creating, deploying, and managing continuous artificial intelligence agents. Prior to 0.6.52, an authenticated user can bypass the SSRF / private-IP protections in SendWebRequestBlock and reach internal network services. _is_ip_blocked() in backend/backend/util/request.py does not normalize IPv4-mapped IPv6 addresses before checking resolved IPs against the blocked IPv4 ranges, and does not block special-use ranges such as 100.64.0.0/10 (CGNAT, RFC 6598). A hostname that resolves to an IPv4-mapped IPv6 address therefore passes validation and the request reaches the embedded internal IPv4 endpoint. This affects all AutoGPT Platform deployments. This vulnerability is fixed in 0.6.52.

AutoGPT is a workflow automation platform for creating, deploying, and managing continuous artificial intelligence agents. Prior to , the `POST /api/integrations/webhooks/{webhook_id}/ping` endpoint fetches the target webhook by primary key alone without verifying that the webhook belongs to the authenticated user. Any authenticated user can supply an arbitrary webhook_id to confirm webhook existence, leak the webhook's OAuth provider type, and in some cases trigger a ping delivery on behalf of another user. This vulnerability is fixed in .

Pagekit CMS 1.0.18 contains a privilege escalation vulnerability that allows authenticated users with the 'user: manage users' permission to escalate privileges by assigning arbitrary custom roles to themselves due to missing authorization checks in UserApiController::saveAction(). Attackers can assign themselves a custom role with the 'system: manage packages' permission and then upload and install a malicious PHP package through the admin package installer to achieve remote code execution.

An observable timing discrepancy in the ASP could allow a privileged attacker to perform a brute-force attack against the hash message authentication code, allowing the input of an arbitrary message, potentially leading to a loss of data integrity.

A privilege escalation vulnerability exists in LXD from 6.0 before 6.9, 5.21.0 before 5.21.5, and 5.0.0 before 5.0.7 regarding the handling of project-restriction policies during snapshot restoration.. An authenticated project operator in a restricted multi-tenant environment can bypass policy restrictions by importing a maliciously crafted instance backup containing restricted configuration keys within a snapshot. When the snapshot is restored, these restricted keys are applied to the live instance without policy validation. Starting the modified instance grants the operator unauthorized host root access.

Kernel driver ProcessMonitorDriver.sys in Safetica's endpoint client x64 , versions 10.5.75.0 and 11.11.4.0, allows unprivileged user to abuse IOCTL path and terminate protected system processes.

Server side template inject (SSTI) in the expression evaluation component in Genshi Template Engine version 0.7.9 allows a remote attacker to achieve remote code execution (RCE) via crafted template expressions.

Docling simplifies document processing by parsing diverse formats and providing integrations with the generative AI ecosystem. Prior to 2.94.0, the HTML backend has unsafe URI and path handling. This vulnerability is fixed in 2.94.0.

An observable timing discrepancy in the ASP could allow a privileged attacker to perform a brute-force attack against the hash message authentication code, allowing arbitrary message input, potentially leading to a loss of data integrity.

Docling simplifies document processing by parsing diverse formats and providing integrations with the generative AI ecosystem. From 2.45.0 until 2.91.0, the METS-GBS backend's XML parsing and the input document format detection lacked security controls. An attacker could craft malicious METS-GBS archives that, when processed, could read sensitive files, exhaust system resources, or cause application crashes. This vulnerability is fixed in 2.91.0.

The default JVM can access files and directories under `/tmp/` including the `$TemporaryDirectory` of other users on the same cloud instance (`/tmp/UserTemporaryFiles/`). The `-init` file for the the JVM initialization exists in the vulnerable directory during the startup of the JVM. An attacker with access to the shared `/tmp/` space can preemptively create or replace `.jar` files or directories (via the `-init` file) that the victim JVM will resolve first in its classpath. By strategically placing a malicious version of a commonly used library (e.g., `commons-io`) in a location that is included in the classpath before the legitimate version, an attacker can cause the JVM to load the malicious class during startup, thereby executing the attacker's code.

Nil-pointer dereference in CreateCustomVolumeFromBackup in LXD up to version 6.8 and 5.21 on Linux allows an authenticated user with can_create_storage_volumes permissions to cause a denial of service via a specially crafted custom-volume backup tarball that omits the expires_at snapshot field.

Broken Access Control in the devLXDInstancePatchHandler component of Canonical LXD allows an untrusted guest to mount, read, and overwrite another guest's custom storage volume via a crafted device PATCH request over /dev/lxd when security.devlxd.management.volumes is enabled.

Kernel software installed and running inside a Host VM may post improper commands to the GPU Firmware to trigger a memory read or write outside the permitted range of memory for the host kernel. Addresses passed to the GPU Firmware can be used by the Firmware for more privileged memory accesses than are permitted by the system.

Unauthenticated remote information disclosure vulnerability in Ollama's model quantization engine allows an attacker to read and exfiltrate the server's heap memory, potentially leading to sensitive data exposure, further compromise, and stealthy persistence.

A web page that contains unusual GPU shader code is loaded into the GPU compiler process and can trigger a write out-of-bounds write crash in the GPU shader compiler library. On certain platforms, when the compiler process has system privileges this could enable further exploits on the device. An edge case using a very small value in GPU shader code can cause a segmentation fault in the GPU shader compiler due to am out-of-bounds write.

Sales Representative SQL Injection in Groundhogg <= 4.5 versions.