CVE-20XX-?
agent?

The internet is broken right now.

68 critical CVEs without a fix in the last 48 hours. Developers are crying quietly into their IDEs. 151 more already have a fix that someone still has to apply.

983 CVEs · updated 2026-11-19 10:52 UTC

Reset all
  1. CVE-2026-58064

    The Redis protocol implementation in Apache bRPC allowed inline command parsing to allocate and copy the full connection buffer before a complete CRLF-terminated command was available. A remote attacker with access to a Redis-protocol bRPC service could send long inline inputs without a terminating CRLF to increase memory usage and CPU copy overhead, potentially causing denial of service. Root Cause: The Redis inline command parser allocated memory based on the current input buffer size and copied the entire buffer before determining the current inline command boundary. The existing redis_max_allocation_size limit was applied to RESP bulk strings and arrays, but not effectively to the inline command path before allocation. Affected Scenarios: Use Apache bRPC server with Redis protocol enabled and reachable by untrusted clients. How to Fix: 1. Upgrade Apache bRPC to version 1.18.0, which fixes this issue. 2. Apply the patch from PR https://github.com/apache/brpc/pull/3363 . Note: The fix applies redis_max_allocation_size to Redis inline command lines and scans for CRLF before allocating parser arena memory. Inline command lines exceeding this limit are rejected.

    Denial of service Library

    7.5 HIGH 0.6% exploit odds No fix listed
  2. CVE-2026-124709

    Issue summary: A QUIC server that rejects a stream opened by the peer may retain that stream's state for the lifetime of the connection if the peer resets the stream before the rejection is signalled, even though the peer's stream credit is returned. Impact summary: A remote client that completes a QUIC handshake can repeatedly open and reset streams, causing server memory to grow without bound until the process runs out of memory and every connection on it is lost. This is a Denial of Service. CWE: CWE-401: Missing Release of Memory after Effective Lifetime Description: QUIC servers track each stream the peer opens. When the incoming stream policy rejects a stream, OpenSSL schedules a STOP_SENDING frame, marks the stream deleted and returns the peer's stream credit. A deleted stream is collected only once its receive part is fully drained, which for a rejected stream means the STOP_SENDING has been acknowledged. If the peer sends RESET_STREAM first, the reset cancels the pending STOP_SENDING, so it is never sent or acknowledged and the stream is never collected. The default configuration reaches this path. Once a connection has a default stream, meaning the application has read from or written to the connection as the tutorial server does, the effective incoming stream policy rejects streams the application does not accept. Applications that set the REJECT policy explicitly are affected too. Each retained stream costs about 300 bytes, and the peer's stream credit window paces how quickly they accumulate. A client that keeps the connection open and keeps opening and resetting streams grows server memory without bound. The severity is Low because growth is slow, needs a long-lived connection, and causes no memory corruption or disclosure. FIPS impact: no The affected code is outside the FIPS module boundary.

    Denial of service Library

    7.5 HIGH 0.6% exploit odds Fix listed
  3. CVE-2026-97043

    Building a malicious package that uses cgo could bypass "#cgo LDFLAGS" validation and cause unexpected code execution at build time, or cause the go command to panic, via crafted quoting in "//go:cgo_ldflag" directives. When "cmd/cgo" generates "_cgo_gotypes.go", it quotes each "#cgo LDFLAGS" argument with strconv.Quote in a "//go:cgo_ldflag" directive, but "cmd/compile" and "cmd/go" parsed those directives by splitting on raw quote characters without handling backslash escapes or unquoting the string. This mismatch allowed disallowed linker flags or unvalidated linker search paths to be smuggled past the go command's flag checks and passed to the external linker. In addition, deduplicating LDFLAGS in the go command could panic with an out-of-bounds slice index when a prefix of the flag list matched near the end of the slice. "cmd/compile" and "cmd/go" now properly handle backslash-escaped quotes and unquote "//go:cgo_*" pragma strings with strconv.Unquote, and the slice bounds check in processCgoOutputs has been fixed.

    Memory safety Library

    8.8 HIGH 0.6% exploit odds No fix listed
  4. CVE-2026-57845

    Hugging Face Datasets before 5.2.0 contains an insecure deserialization vulnerability that allows attackers to execute arbitrary Python code by crafting a malicious pickle file loaded through the deprecated pandas packaged builder. Attackers can supply a pickle-backed dataset that is processed by Pandas._generate_tables via pandas.read_pickle() without validation, so when a victim calls datasets.load_dataset("pandas", data_files=...), the embedded __reduce__ payload executes code with the victim process's privileges.

    Code execution Library

    8.7 HIGH 0.6% exploit odds Fix listed
  5. CVE-2026-121953

    FreeRDP before 3.32.2 has a use-after-free in the client's rdpgfx dynamic virtual channel: a duplicate-name channel with a new ChannelId overwrites the listener's single callback slot, which rdpgfx_on_close frees without clearing. A malicious RDP server can open and close such a channel, then send a PDU on the original channel to crash the client or potentially execute code.

    Memory safety Library Freerdp

    9.3 CRITICAL 0.6% exploit odds Fix listed
  6. CVE-2026-121611

    A flaw was found in GStreamer's gst-plugins-bad package. A use of uninitialized memory vulnerability exists in the H.265 (HEVC) bitstream parser when processing Video Parameter Sets that declare more than one HRD (Hypothetical Reference Decoder) parameter set. A remote attacker could exploit this by providing a crafted H.265 media file or stream that, when processed by an application using GStreamer, causes unpredictable parser behavior and may crash the application. This can lead to a denial of service for any application using GStreamer to play back H.265 content.

    Denial of service Library

    6.5 MEDIUM 0.6% exploit odds No fix listed
  7. CVE-2026-119843

    A flaw was found in GStreamer's gst-plugins-good package. The FLV media demuxer does not properly validate buffer sizes and data availability when processing enhanced audio and video tags in malformed or truncated FLV streams. A remote attacker could exploit this vulnerability by providing a specially crafted FLV media file that, when opened or streamed by a GStreamer-based application, triggers an assertion failure or a process crash, leading to denial of service.

    Denial of service Library

    6.5 MEDIUM 0.6% exploit odds No fix listed
  8. CVE-2026-119777

    A flaw was found in GStreamer's gst-plugins-bad H.266/VVC parser. When processing a malformed H.266 media file or stream, an integer overflow while parsing profile, tier, and level fields can lead to a stack-based buffer overflow. An attacker could cause an application to crash and may be able to execute code in the context of the affected application.

    Memory safety Library

    8.8 HIGH 0.6% exploit odds No fix listed
  9. CVE-2026-85783

    Crypt::Age versions before 0.004 for Perl allow CPU exhaustion via an unbounded number of recipient stanzas, each costing an X25519 scalar multiplication in unwrap_file_key. The header parser accepts any number of recipient stanzas, and unwrap_file_key tries every identity against every X25519 stanza. Each attempt runs a scalar multiplication, an HKDF and a ChaCha20-Poly1305 open. The header MAC cannot be checked first, because its key comes from the file key, which only a successful unwrap produces. Each stanza is 98 bytes of attacker-chosen header and buys one scalar multiplication per identity held, and the file is rejected only after every one of them has run. A caller reaches this through decrypt, decrypt_file or decrypt_filehandle whenever it decrypts a file it did not write.

    Denial of service Library

    7.5 HIGH 0.6% exploit odds No fix listed
  10. CVE-2026-118485

    Axel Download Accelerator 2.17.14 and earlier performs URL percent-decoding (http_decode()) on the output filename after the URL has been parsed and the filename component extracted. This allows an attacker to embed %2e%2e%2f (URL-encoded ../) sequences in the URL path. These sequences survive the URL parser intact (they contain no literal / characters) but are decoded into directory traversal sequences before the output file is opened for writing. As a result, a crafted URL can cause axel to write downloaded content to arbitrary filesystem paths relative to the current working directory.

    Disclosure Library

    7.0 HIGH 0.6% exploit odds Fix listed
  11. CVE-2026-101259

    Generation of Error Message Containing Sensitive Information vulnerability in Apache Camel CoAP component. This issue affects Apache Camel: from 4.0.0 before 4.18.5, from 4.19.0 before 4.22.2. When the CoAP consumer could not build its response once the route had completed, CamelCoapResource caught the resulting exception and sent its message back to the CoAP client as the payload of a 5.00 Internal Server Error response. The typical case is a type conversion failure: the consumer converts the final message body to the byte[] payload, and the CamelCoapMaxAge and CamelCoapETag headers to their types, after the route has returned, and therefore outside the route's error handler. On the affected releases the message of such a failure names the class of the value being converted, carries the value's string form, and ends with the class name and message of the underlying cause, so a remote CoAP client could read internal detail about the deployment: class and package names, the string form of the value, such as the path of a file body, and whatever the cause carries, which in practice can include paths, host names and connection detail of a backend. The consumer had no option to suppress this, unlike the HTTP consumers, which offer a muteException option for that purpose. The exception raised by a failing route was not sent: the consumer did not check whether the exchange had failed, and answered a route failure with 2.05 Content and whatever body the message held at that point. Reaching the disclosure therefore requires a route whose completed response a remote client can make unconvertible, which depends on how the route is written. Users are recommended to upgrade to version 4.23.0, which fixes the issue. If users are on the 4.22.x releases stream, then they are suggested to upgrade to 4.22.2. If users are on the 4.18.x releases stream, then they are suggested to upgrade to 4.18.5. The 4.14.x releases stream and earlier, and the 4.15.x to 4.17.x and 4.19.x to 4.21.x release lines, reached end of life before this issue was resolved and will not receive a fix, so deployments on them must move to a supported release. The upgrade is enough on its own: the new muteException consumer option defaults to true on all three fixed release lines, so the error response carries no payload without any configuration change. For deployments that cannot upgrade immediately, make sure the response can always be built once the route has completed: convert the body to byte[] or String as the last step of each route that consumes from a coap endpoint, for example with convertBodyTo, give the CamelCoapMaxAge and CamelCoapETag headers only values of their own types, and add an onException handler that marks the failure as handled and sets the response body explicitly, so that a conversion failure is raised and replaced inside the route rather than in the consumer.

    Disclosure Library Apache Camel

    5.3 MEDIUM 0.6% exploit odds Fix listed
  12. CVE-2026-126303

    NLTK before 3.10.4 contains an uncontrolled resource consumption vulnerability in SteppingRecursiveDescentParser.parse() that lacks any max_time deadline or step limit. Attackers influencing the grammar can supply an ambiguous self-embedding grammar and short token input to force unbounded top-down search, pinning CPU cores and exhausting worker pools.

    Denial of service Library

    8.7 HIGH 0.6% exploit odds Fix listed
  13. CVE-2026-121960

    FreeRDP before 3.32.2 contains a denial of service vulnerability in the NLA-enabled shadow server when a CredSSP TSRequest omits negoTokens. Unauthenticated attackers can send a crafted seven-byte TSRequest containing only the version field to trigger a fatal WINPR_ASSERT and abort the server process.

    Denial of service Library Freerdp

    8.7 HIGH 0.6% exploit odds Fix listed
  14. CVE-2026-54140

    OpenHuman is an open-source agent harness available as a desktop application and Rust library. Before 0.57.13, OpenHuman places names, descriptions, titles, and parameter definitions returned by configured external MCP servers into the agent model context without neutralizing instructions or bounding the descriptive fields. The McpRemoteTool records produced by list_tools preserve this remote content, while tool filtering considers the tool name and the prompt-injection detector screens user messages rather than MCP definitions. A malicious or compromised MCP server can therefore return a poisoned description that is evaluated on every agent turn and may direct the model to invoke other available tools, disclose accessible data, or perform unauthorized actions without a separate malicious user message. The issue affects users who configure an attacker-controlled or compromised MCP server. This issue is fixed in version 0.57.13.

    Injection Library

    8.3 HIGH 0.6% exploit odds Fix listed
  15. CVE-2026-111959

    Out-of-bounds read vulnerability in Apache Arrow Rust (`arrow` crate), from version 33.0.0 through to versions 57.3.1, 58.4.0, 59.3.1 and 60.0.0. Improper validation of RunEndEncoded arrays in the ArrayData interface can allow reading data beyond the bounds of the child values array. Users are recommended to upgrade to versions 60.1.0, 59.3.2, 58.4.1, 57.3.2 which fixes the issue.

    Memory safety Library

    9.1 CRITICAL 0.6% exploit odds Fix listed
  16. CVE-2026-116570

    FreeSWITCH is a Software Defined Telecom Stack enabling the digital transformation from proprietary telecom switches to a software implementation that runs on any commodity hardware. Prior to 1.11.2, http_directory_auth() rebuilt an expected user@domain:password credential and base64-encoded it into an 80-byte buffer using an encoder without an output-length argument. A configured credential longer than 57 characters overflowed the stack before the supplied HTTP Basic password was compared. An unauthenticated client only needed to name an existing directory user on an exposed mod_xml_rpc HTTP listener; the module is built but not loaded by the stock configuration. The overflow can corrupt the stack and crash FreeSWITCH, ending every active call, although the overrun bytes come from the configured credential and remote code execution has not been demonstrated. This issue is fixed in version 1.11.2.

    Memory safety Library Freeswitch

    7.5 HIGH 0.6% exploit odds Fix listed
  17. CVE-2026-65964

    Pi-hole FTL is the core engine of the Pi-hole network-level advertisement and tracker blocker. Prior to 6.7, log_http_access in src/webserver/webserver.c writes attacker-controlled User-Agent content to the access log without output neutralization when debug.api is enabled. If an administrator configures files.log.webserver inside webserver.paths.webroot with a filename matching lua_server_page_pattern, an unauthenticated request can plant Lua markup in that log and a later request can cause CivetWeb to evaluate the log as a Lua server page. The resulting commands run as the pihole operating-system user and can access or modify pihole-owned configuration, credentials, databases, logs, and rendered interface content. When the log is not served through the Lua handler, the same unneutralized writer still permits log forgery and downstream log-parser injection. This issue is fixed in version 6.7.

    Injection Library

    6.6 MEDIUM 0.6% exploit odds Fix listed
  18. CVE-2026-59171

    JSON-java is a reference implementation of a JSON package in Java. From version 20150729 through 20260522, JSONTokener.nextSimpleValue can accumulate unbounded numeric tokens before JSONObject.stringToNumber passes them to BigInteger or BigDecimal conversion without a length limit. An application that parses attacker-controlled JSON or XML through JSONObject, JSONArray, or XML.toJSONObject can spend super-linear CPU time converting a sufficiently long numeric literal, blocking request-handling threads and causing denial of service. Strict mode and keep-strings settings in JSONParserConfiguration do not prevent the affected conversion path. Version 20260719 limits numeric conversion length.

    Denial of service Library

    5.9 MEDIUM 0.6% exploit odds Fix listed
  19. CVE-2026-119578

    MONAI from 1.6.0 before 1.6.1 contains a remote code execution vulnerability in Auto3DSeg algo_from_json() that allows attackers to run Python code via a crafted work directory's algo_object.json. Attackers can ship a malicious algorithm_templates module plus a _target_ name that is imported and instantiated when victims call AutoRunner or import_bundle_algo_history, executing code under default configuration.

    Code execution Library

    8.5 HIGH 0.6% exploit odds Fix listed
  20. CVE-2026-86146

    Inconsistent interpretation of HTTP requests (HTTP request/response smuggling) in the bundled HTTP/1.1 parser of Apache bRPC (version < 1.18.0) on all platforms allows remote attackers to smuggle requests and inject headers when bRPC is deployed behind a proxy or load balancer that parses the same bytes differently. Root Cause: The bRPC fork of http-parser accepts and acts on several HTTP/1.1 header forms that a conforming front-end either rejects or interprets differently, so the framing or the header set of a message can differ between the front-end and bRPC: 1. A space before the colon of framing headers (e.g. Content-Length : 5) is accepted; the value still drives bRPC's framing while the reported field name contains the trailing space. 2. Interior spaces inside a Content-Length value (e.g. Content-Length: 1 3) are skipped, so bRPC reads 13 where a front-end reads 1 or rejects the message (same class as CVE-2022-32213 in Node.js). 3. Obsolete line folding (obs-fold, RFC 7230 3.2.4) of Content-Length and Transfer-Encoding values is accepted, so a folded framing value drives bRPC's framing while a front-end that does not unfold sees a different message. 4. Chunked trailer fields are fed to the same header callbacks as the header section and are merged into the header map like real headers. A front-end only inspects the header section, so any chunked request can smuggle fields such as Authorization or X-Forwarded-For past it. 5. Transfer-Encoding lists with optional whitespace after a comma (e.g. Transfer-Encoding: gzip, chunked) are not recognized as chunked: such requests are rejected, and responses are read until connection close instead of as chunked, causing response desync. Affected Scenarios: bRPC server or client communicating over HTTP/1.1 with a front proxy, load balancer, or any peer that parses the message independently. An attacker able to send requests through the front-end can desynchronize the two parsers to smuggle requests, forge headers (e.g. Authorization, X-Forwarded-For) via chunked trailers, or desynchronize responses. How to Fix: (Choose one of the following options) 1. Upgrade bRPC to version 1.18.0, which fixes this issue. 2. Apply this patch: https://github.com/apache/brpc/pull/3521 Note: After applying the fix, the four header forms above are rejected by default and chunked trailers are dropped instead of merged into the header map. Two gflags are provided for compatibility: http_allow_obs_fold=true restores acceptance of folded Content-Length/Transfer-Encoding values, and http_strict_header_token=true additionally rejects spaces inside any header field name (not only the ones that decide framing). Legitimate clients that do not use these obsolete forms are unaffected.

    Other Library

    7.5 HIGH 0.5% exploit odds No fix listed
  21. CVE-2026-82230

    Uncontrolled recursion in the mcpack2pb component in Apache bRPC (version < 1.18.0) on all platforms allows remote attackers to make the server crash via sending deeply nested mcpack data. Root Cause: The bRPC mcpack2pb parser (used to serve nshead_mcpack / ubrpc / public_pbrpc requests) recursively parses nested objects and arrays without any depth limit. For a protobuf message that references itself (e.g. a repeated message field), the generated parser recurses once per nesting level of the incoming mcpack object. A deeply nested request makes the recursion grow unbounded, overflow the stack and crash the process. Affected Scenarios: Use bRPC server with a recursive-schema protobuf message to serve nshead_mcpack / ubrpc / public_pbrpc requests from untrusted network. How to Fix: (Choose one of the following options) 1. Upgrade bRPC to version 1.18.0, which fixes this issue. 2. Apply this patch: https://github.com/apache/brpc/pull/3499 Note: No matter which option you choose, you should know that the fix enforces the documented parsing depth limit MAX_DEPTH (128) on the parse path, the same limit the serializer already applies. If your requests contain mcpack messages nested deeper than 128 levels, the request will be failed after applying the fix.

    Denial of service Library

    7.5 HIGH 0.5% exploit odds No fix listed
  22. CVE-2026-107891

    musl libc before 1.2.7, when an untrusted regular expression is used with regcomp, has multiple integer overflows, possibly leading to incorrect allocation sizes and out-of-bounds heap writes.

    Memory safety Library

    8.1 HIGH 0.5% exploit odds Fix listed
  23. CVE-2026-112899

    Pillow is a Python imaging library. In versions 2.3.0 through 12.3.0, EPS rendering does not validate the scaled image dimensions before invoking Ghostscript. Applications that process untrusted EPS or PostScript files using an application-selected scale greater than one can incur excessive Ghostscript CPU and temporary disk consumption. In versions with decompression-bomb protection, the scaled output is checked only after Ghostscript has completed rendering, allowing resource consumption beyond the configured pixel limit. The amplification depends on the scale selected by the application, not an attacker-controlled scale value. This issue is fixed in version 13.0.0.

    Denial of service Library Python Pillow

    6.3 MEDIUM 0.5% exploit odds Fix listed
  24. CVE-2026-128081

    AutoMapper is a convention-based object-object mapper in .NET. Prior to 15.1.4 and from 16.0.0 until 16.3.0, AutoMapper can recurse without an effective MaxDepth limit when mapping polymorphic graphs through Include or IncludeBase, runtime-typed object or interface members, nested Dictionary<string, object> or ExpandoObject sources, or recursive ConvertUsing type converters. Deep attacker-influenced graphs on these paths can throw an uncatchable StackOverflowException and terminate the process. This issue is fixed in versions 15.1.4 and 16.3.0.

    Denial of service Library LuckyPennySoftware AutoMapper

    7.5 HIGH 0.5% exploit odds Fix listed
  25. CVE-2026-77473

    AsyncSSH is a Python package which provides an asynchronous client and server implementation of the SSHv2 protocol on top of the Python asyncio framework. Prior to 2.24.1, recursive remote-to-local SFTP copies in SFTPClient._copy() in asyncssh/sftp.py, reached through get(), mget(), and copy(), trust filenames returned by a remote server's directory enumeration. A malicious or compromised SFTP server can return an absolute filename or a filename containing path separators and parent-directory traversal, causing destination path construction to escape the requested local directory and write attacker-selected content to locations writable by the client process. This issue is fixed in version 2.24.1.

    Disclosure Library

    6.5 MEDIUM 0.5% exploit odds Fix listed
  26. CVE-2026-114613

    MessagePack is the serializer implementation for Python msgpack.org. Prior to 1.2.3, the msgpack-python C extension (msgpack._cmsgpack) leaks a decoded map value when deserialization rejects a key with strict_map_key=True (the default). Repeated malformed inputs can cause unbounded process memory growth and denial of service in applications that deserialize attacker-controlled MessagePack data. Both unpackb() and the streaming Unpacker API are affected. The pure-Python fallback does not exhibit this leak in the reported tests. This issue is fixed in version 1.2.3.

    Denial of service Library

    7.5 HIGH 0.5% exploit odds Fix listed
  27. CVE-2026-97038

    Building a malicious package containing ".syso" files or cgo linker inputs could inject arbitrary linker flags or execute arbitrary code at link time. Linker input files—including ".syso" files, direct object or library inputs in "#cgo LDFLAGS", and libraries in "-L" directories—were passed to the external linker without validating their format or checking for embedded linker options. Across platforms, this allowed executing GNU ld implicit linker scripts disguised as libraries or ".syso" files (or loaded via "-l:<filename>" or "-l" containing path separators), injecting unallowlisted linker flags via PE/COFF ".drectve" sections on Windows or Mach-O "LC_LINKER_OPTION" load commands on Darwin, or invoking LTO plugins via GCC or LLVM LTO intermediate representation sections in ELF, PE/COFF, or Mach-O objects. The go command now verifies that ".syso" files, direct linker inputs, and libraries in "-L" directories are valid binary objects or archives that do not contain unallowlisted ".drectve" directives, "LC_LINKER_OPTION" load commands, or LTO IR sections, and disallows "-l:<filename>" and ":", "=", or path separators in "-l" flags.

    Code execution Library

    8.8 HIGH 0.5% exploit odds No fix listed
  28. CVE-2026-97033

    Building a malicious package that uses cgo could cause arbitrary code execution at build time through several compiler and linker flags permitted in "#cgo" directives. Specifically, "--sysroot" in GCC and Clang relocates compiler subprograms, spec files, and CRT startup objects to an attacker-controlled directory; "-target" in Clang falls back to executing target helper tools from the current working directory when not found in PATH; "-isysroot" and "-Wl,-syslibroot" can pass a regular file on disk to GNU ld.bfd, which interprets it as a positional linker script; and "-x" permitted specifying arbitrary compiler input languages beyond those supported by cgo. "--sysroot" and "-target" are now disallowed in "#cgo" flags, "-isysroot" and "-Wl,-syslibroot" are verified not to point to regular files, and "-x" is restricted to the C, C++, Objective-C, and assembler languages supported by cgo.

    Code execution Library

    8.8 HIGH 0.5% exploit odds No fix listed
  29. CVE-2026-97035

    When using Clang or GCC, building a malicious package that uses cgo could cause unexpected code execution at build time or write files to arbitrary directories via "-gen-*" or "-gnat*" flags in "#cgo" directives. The "-g*" wildcard in the compiler and linker flag allowlists previously matched Clang's two-argument "-gen-cdb-fragment-path <dir>" flag and GCC's two-argument "-gnatO <file>" flag, permitting compilation database fragments and sidecar output files to be written to arbitrary directories and desynchronizing flag argument parsing between the go command and the compiler driver so that subsequent arguments could be passed as unvalidated positional inputs to the linker. "-gen-*" and "-gnat*" flags are now disallowed in "#cgo" compiler and linker flags.

    Other Library

    8.8 HIGH 0.5% exploit odds No fix listed
  30. CVE-2026-111771

    The Model Context Protocol TypeScript SDK provides TypeScript client and server implementations of the Model Context Protocol. From 1.6.0 until 1.32.0 in @modelcontextprotocol/sdk, from 2.0.0 until 2.3.0 in @modelcontextprotocol/server, from 2.0.0 until 2.0.2 in @modelcontextprotocol/express, and from 2.0.0 until 2.3.1 in @modelcontextprotocol/server-legacy, server-side bearer authentication could accept a valid access token issued for a different service because the SDK did not enforce the token audience. The affected surface is server-side HTTP bearer authentication. The trigger is a valid token whose audience names another service. The authorization server must issue tokens for multiple services, and the application's verifyAccessToken() must not independently validate aud. Applications already validating aud inside verifyAccessToken() are not affected. An attacker with a token for another service could authenticate to the MCP server and obtain the confidentiality and integrity capabilities granted to an authenticated caller. Upgrading alone leaves the new audience validation disabled. Operators must enable audience validation and provide the token audience together. Enabling only the expected audience check causes every request to receive HTTP 401. This issue is fixed in @modelcontextprotocol/sdk 1.32.0, @modelcontextprotocol/server 2.3.0, @modelcontextprotocol/express 2.0.2, and @modelcontextprotocol/server-legacy 2.3.1.

    Auth & access Library

    6.8 MEDIUM 0.5% exploit odds Fix listed
  31. CVE-2026-126190

    Soup Sieve is a CSS selector library designed to be used with Beautiful Soup 4. Prior to 4.0, Soup Sieve repeatedly walks the same ancestors while matching descendant combinators, and each additional descendant combinator raises the polynomial degree of the work on deeply nested attacker-influenced HTML or XML. Applications using selectors such as section div div through select(), select_one(), or Beautiful Soup CSS APIs against untrusted documents can have a worker occupied for seconds or minutes, while trusted-only document processing is not affected. Selectors compiled with NOCACHE continue to use the old uncached behavior. This issue is fixed in version 4.0.

    Other Library

    7.5 HIGH 0.5% exploit odds Fix listed
  32. CVE-2026-116782

    FreeSWITCH is a Software Defined Telecom Stack enabling the digital transformation from proprietary telecom switches to a software implementation that runs on any commodity hardware. Prior to 1.11.2, rtmp_i_connect() passed the first RTMP connect argument to amf0_object_get() without checking whether the AMF value was an object. A string or number in place of an AMF object caused union bytes from the message to be interpreted as node pointers and dereferenced. Any client with network access to an enabled mod_rtmp listener could send the malformed connect command before login; the module is disabled in the stock build and configuration. The invalid read reliably crashes FreeSWITCH and ends every active call, while no write, memory disclosure, or code execution has been demonstrated. This issue is fixed in version 1.11.2.

    Memory safety Library Freeswitch

    7.5 HIGH 0.5% exploit odds Fix listed
  33. CVE-2026-126994

    A vulnerability has been found in braindecode up to 1.8.1. Affected is the function resolve_type_kwargs of the file braindecode/models/util.py of the component Model Config Deserialization. Such manipulation of the argument Activation leads to use of externally-controlled input to select classes or code. The attack can be launched remotely. The exploit has been disclosed to the public and may be used. Upgrading to version 2.0.0 is able to address this issue. The name of the patch is 05b764d93b9a50400f523206fbe03d73d0cf5edb. The affected component should be upgraded.

    Code execution Library

    2.1 LOW 0.5% exploit odds Fix listed
  34. CVE-2026-115557

    Python has added a denylist for known compromised certificates to the "ssl" module to ensure user safety even without certificate revocation configured. Known compromised certificates will automatically be rejected regardless of the SSLContext policy configuration.

    Crypto Library

    9.4 CRITICAL 0.5% exploit odds Fix listed
  35. CVE-2026-123261

    CAI Content Credentials is affected by an Improper Verification of Cryptographic Signature vulnerability that could result in a Security feature bypass. An attacker could leverage this vulnerability to bypass security measures and gain unauthorized write access. Exploit depends on conditions beyond the attacker's control. Exploitation of this issue does not require user interaction.

    Crypto Library

    5.9 MEDIUM 0.5% exploit odds No fix listed
  36. CVE-2026-123262

    CAI Content Credentials is affected by an Improper Verification of Cryptographic Signature vulnerability that could result in a Security feature bypass. An attacker could leverage this vulnerability to bypass security measures and gain unauthorized write access. Exploit depends on conditions beyond the attacker's control. Exploitation of this issue does not require user interaction.

    Crypto Library

    5.9 MEDIUM 0.5% exploit odds No fix listed
  37. CVE-2026-113572

    CAI Content Credentials is affected by a Use of a Broken or Risky Cryptographic Algorithm vulnerability that could result in a Security feature bypass. An attacker could leverage this vulnerability to bypass security measures and gain unauthorized write access. Exploit depends on conditions beyond the attacker's control. Exploitation of this issue does not require user interaction.

    Crypto Library

    5.9 MEDIUM 0.5% exploit odds No fix listed
  38. CVE-2026-128089

    dnspython is a DNS toolkit for Python. Prior to version 2.9.2, the from_text() parser for WKS records does not require the service number to be between 0 and 65535 before using it as a bitmap offset. An attacker-controlled zone file can supply a larger number, causing the service bitmap to grow to an extremely large size through inefficient repeated extension, consuming excessive memory and CPU and potentially terminating the process. Applications are affected only when they parse zone-file text whose record type and rdata are attacker controlled. This issue is fixed in version 2.9.2.

    Denial of service Library Rthalley Dnspython

    7.5 HIGH 0.5% exploit odds Fix listed
  39. CVE-2026-126366

    MessagePack for Ruby is an implementation of the MessagePack binary serialization format. Prior to 1.8.7, the Java-platform msgpack gem for JRuby allocated arrays and byte buffers directly from attacker-controlled array32, str32, and bin32 declared sizes before confirming that corresponding elements or bytes were available. A short MessagePack header could therefore trigger a very large allocation or sustained heap growth when an application decoded untrusted input through MessagePack.unpack or MessagePack::Unpacker, potentially exhausting the shared JVM heap. map32 and declared counts at or above 2^31 did not use the affected allocation behavior. This issue is fixed in version 1.8.7.

    Denial of service Library

    7.5 HIGH 0.5% exploit odds Fix listed
  40. CVE-2026-16550

    An external control of file name or path vulnerability in the Media Parser in Synology DiskStation Manager (DSM) before 7.3.2-86009-5 and 7.4.1-90080-1 allows remote authenticated users to write arbitrary files.

    Other Library

    6.5 MEDIUM 0.5% exploit odds No fix listed
  41. CVE-2026-119195

    pyasn1 is a generic ASN.1 library for Python. Prior to 0.6.5, constructed string and indefinite-length ANY decoding repeatedly concatenates each fragment onto a growing bytes value. Attacker-controlled BER or CER data can use many constructed OCTET STRING or character-string fragments. DER rejects constructed OCTET STRING without a schema, but accepts its definite-length form when asn1Spec selects univ.OctetString; it also accepts constructed character strings such as UTF8String without a schema, even though those encodings are not valid DER. The separate indefinite-length ANY path is reachable in BER and CER when asn1Spec selects univ.Any; DER rejects indefinite lengths. OctetStringPayloadDecoder uses growing header concatenation and AnyPayloadDecoder.indefLenValueDecoder uses growing chunk concatenation, repeatedly copying an increasing prefix and causing quadratic work. A sufficiently fragmented value can occupy decoding workers and cause CPU exhaustion without confidentiality or integrity impact. This issue is fixed in version 0.6.5.

    Denial of service Library

    7.5 HIGH 0.5% exploit odds Fix listed
  42. CVE-2026-121707

    Effect is a TypeScript framework that consists of several packages that work together to help build TypeScript applications. Prior to 3.22.1, Schema.TemplateLiteral schemas containing at least two Schema.String spans compile each span to the ambiguous lazy wildcard [\s\S]*?. Decoding a non-matching untrusted string causes cubic regular expression backtracking, which can block the JavaScript event loop. The issue affects Schema.decodeUnknown, Schema.decodeUnknownSync, Schema.is, and consumers such as HttpApi and @effect/rpc payload decoding. This issue is fixed in version 3.22.1.

    Denial of service Library Effect-TS Effect

    8.7 HIGH 0.5% exploit odds Fix listed
  43. CVE-2026-54142

    OpenHuman is an open-source agent harness available as a desktop application and Rust library. Before 0.57.13, background synchronization for connected services such as email, Slack, and Notion stores raw external content and later places it in the model context of autonomous background agents without neutralizing embedded instructions. ApprovalGate treats tool calls without an interactive chat context as pre-authorized automation, so attacker-authored content processed during a background tick can influence the model to invoke Composio actions without presenting an approval request. An attacker who can deliver content to a connected account can use this path to send or forward data, post messages, create or modify third-party content, or disclose information from OpenHuman memory and integrations. The issue requires an active integration and successful model interpretation of the injected instructions. This issue is fixed in version 0.57.13.

    Injection Library

    8.7 HIGH 0.5% exploit odds Fix listed
  44. CVE-2026-103582

    Unbounded memory allocation and missing frame size validation in Apache NMS ActiveMQ (the .NET OpenWire client, package Apache.NMS.ActiveMQ) before 2.3.0. The OpenWire unmarshalling code trusted the frame size and the length and count fields read from the network. It did not enforce a maximum frame size, and it did not check that decoded lengths fit within the frame being read. A malicious or compromised broker, or an attacker able to intercept and modify an unencrypted connection, could send a crafted OpenWire frame with an oversized or negative length. This could cause the client to attempt very large memory allocations, leading to excessive memory consumption, an OutOfMemoryException or an uncaught exception, and a denial of service of the client application. Users are recommended to upgrade to version 2.3.0, which enforces a configurable maximum frame size and validates lengths against the size of the frame being read.

    Memory safety Library

    7.5 HIGH 0.5% exploit odds Fix listed
  45. CVE-2026-125712

    A flaw was found in GStreamer's gst-plugins-base RTCP SDES packet parser. A remote sender able to reach an active media session can supply a malformed SDES entry whose declared length exceeds the received packet. Missing bounds validation can cause the receiver to read beyond the packet buffer, potentially crashing the application or exposing a limited amount of process memory.

    Memory safety Library

    7.5 HIGH 0.5% exploit odds No fix listed
  46. CVE-2026-115774

    Twisted is an event-based framework for internet applications, supporting Python 3.6+. Prior to 26.10.0rc1, twisted.mail.imap4.IMAP4Server reaches twisted.mail.imap4.parseNestedParens with a negative literal size in an authenticated SEARCH command after the client enters the selected mailbox state. The negative size moves the parser cursor below zero, and Python's negative indexing prevents the synchronous scanning loop from terminating. The loop freezes the single-threaded reactor, stopping other IMAP connections and any SMTP or POP3 services hosted in the same process until restart. This issue is fixed in version 26.10.0rc1.

    Denial of service Library

    4.9 MEDIUM 0.5% exploit odds Fix listed
  47. CVE-2026-129266

    External control of a file path in Amazon Deep Java Library (DJL) versions 0.7.0 through 0.38.0 might allow a remote unauthenticated third party to read arbitrary files that the library process can access. To do this, the third party supplies a crafted model archive. To remediate this issue, users should upgrade to version 0.39.0 or later.

    Disclosure Library

    8.7 HIGH 0.5% exploit odds Fix listed
  48. CVE-2026-126360

    OpenTelemetry Python provides the Python API and SDK for OpenTelemetry. Prior to 1.45.2, TraceContextTextMapPropagator.extract() in opentelemetry-api can perform quadratic-time regular-expression backtracking on a malformed traceparent value containing a trace-context prefix, a long whitespace sequence, and an embedded newline. Exploitation requires an attacker-controlled carrier that preserves embedded newlines and permits a sufficiently large value, because conforming HTTP header values cannot contain embedded newlines. JSON-based carriers, including applicable MCP request metadata integrations, can preserve the required newline depending on the SDK, instrumentation, and message-size limits. Parsing on an asynchronous server's event-loop thread can consume excessive CPU and delay unrelated requests, while the advisory does not establish disclosure, modification, or code execution. This issue is fixed in version 1.45.2.

    Denial of service Library

    5.3 MEDIUM 0.5% exploit odds Fix listed
  49. CVE-2026-107360

    A flaw was found in rubygem-safemode. Compound assignment expressions in sandboxed code are evaluated without mediating method calls through the Jail allow-list. An authenticated attacker who can supply template content can invoke arbitrary no-argument methods on objects exposed to the template renderer without Jail mediation, including destructive operations such as host deletion and disclosure of non-allow-listed object state.

    Other Library

    9.6 CRITICAL 0.5% exploit odds No fix listed
  50. CVE-2026-115533

    The LeRobot library is vulnerable to arbitrary code execution via unsafe dynamic class instantiation in the from_pretrained() method and related paths. When loading a policy's processor configuration from a repository, the library uses the class field within the JSON configuration to dynamically resolve and instantiate Python classes using importlib.import_module() and getattr().

    Code execution Library

    8.8 HIGH 0.5% exploit odds No fix listed