| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| vLLM through 0.29.0 fetches and fully materializes remote or inline media before enforcing its documented media controls (the VLLM_MAX_AUDIO_CLIP_FILESIZE_MB compressed-audio size cap, default 25 MB, and the per-modality --limit-mm-per-prompt item limits). Across four ingress paths — the shared media-acquisition layer (HTTPConnection.get_bytes()/async_get_bytes()), the chat completions audio_url/base64 path, the batch speech runner, and the Rust frontend POST /tokenize route — the server reads the entire HTTP response body, base64-decodes the inline payload, or spawns one fetch/decode task per media part, and only then applies the limit (or, on some paths, never applies it). A remote attacker can therefore cause the API server or batch-runner process to allocate memory and consume outbound bandwidth proportional to an attacker-chosen body size or media item count before the request is rejected, resulting in pre-inference memory and bandwidth exhaustion (denial of service). The chat and batch surfaces require an API key when one is configured; the Rust frontend /tokenize route is unauthenticated by design. There is no code execution or data disclosure impact. |
| An unauthenticated client can drain the RTSP server's packet pool with a couple of dozen requests
that carry a Session header the parser cannot convert.
The Session branch returns the raw NetX error code instead of an RTSP status code:
```c
/* addons/rtsp/nx_rtsp_server.c:2754 */
status = _nx_utility_string_to_uint(field_value_ptr, field_value_length, &session_id);
if (status)
{
return(status); /* NX_INVALID_PARAMETERS / NX_SIZE_ERROR / NX_OVERFLOW */
}
```
Every other branch of the same function maps its failure to an RTSP status first. The CSeq branch
eighteen lines earlier does exactly that (line 2736 returns NX_RTSP_STATUS_CODE_BAD_REQUEST). The
raw code then reaches `_nx_rtsp_server_error_response_send` (nx_rtsp_server.c:1234), which does not
recognise it, takes a path that returns without releasing the response packet it already allocated,
and the block never goes back to the pool.
Six requests with an empty Session header against a 22 packet pool:
```
valid requests: after request 6: pool available = 21, AFTER = 22 / 22
malformed requests: after request 6: pool available = 16, AFTER = 17 / 22
```
One block per request, not returned when the client disconnects. Twenty six requests take the pool
to zero and the server starts failing allocations, after which it serves nobody. If the pool is
shared with the rest of the application, as it is in the shipped sample, the rest of the stack
stops with it.
Convert the `_nx_utility_string_to_uint` failure in the Session branch into
NX_RTSP_STATUS_CODE_BAD_REQUEST the way the CSeq branch does, and release the response packet on
every exit path of `_nx_rtsp_server_error_response_send`. |
| Uncontrolled Resource Consumption (CWE-400) in Elasticsearch can lead denial of service via Excessive Allocation (CAPEC-130) |
| Uncontrolled Resource Consumption (CWE-400) in Elasticsearch can lead denial of service via Excessive Allocation (CAPEC-130) |
| Zammad is a web based open source helpdesk/customer support system. Prior to 7.0.2, summary An unauthenticated request to POST /api/v1/import/otrs/import_check blocks a Zammad request worker for roughly two minutes. The import_check and import_status actions are missing the setup_done_response guard that other actions in the same controller carry, so they execute on fully set-up production instances. The action enters a retry loop against a blank OTRS endpoint, sleeping for 30 s + 45 s between attempts. Impact An unauthenticated remote attacker denies service to a production Zammad instance. Each request costs almost nothing and forces ~115 seconds of server-side blocking. A few requests per second saturate the Puma worker pool. The condition persists as long as the traffic continues. No account, valid import configuration, or target knowledge beyond the hostname is needed. CSRF token is trivially obtained from any prior GET response. This issue is fixed in version 7.0.2. |
| The brace-expansion library generates arbitrary strings containing a common prefix and suffix. Prior to 1.1.21, 2.1.7, 3.0.9, and 5.0.12, the expand function handles untrusted {a},b}-shaped patterns with many trailing closing braces by restarting its scan once for each trailing closing brace. The successive full-input rescans with linear working-string growth cause quadratic CPU time and memory pressure that can block the Node.js event loop. The process eventually recovers, making the impact a recoverable CPU denial of service. This issue is fixed in versions 1.1.21, 2.1.7, 3.0.9, and 5.0.12. |
| Uncontrolled Resource Consumption (CWE-400) in Elasticsearch can lead denial of service via Excessive Allocation (CAPEC-130) |
| Uncontrolled Resource Consumption (CWE-400) in Kibana can lead denial of service via Excessive Allocation (CAPEC-130) |
| Uncontrolled Resource Consumption (CWE-400) in Elasticsearch can lead denial of service via Excessive Allocation (CAPEC-130) |
| Uncontrolled Resource Consumption (CWE-400) in Elasticsearch can lead to denial of service via Excessive Allocation (CAPEC-130). |
| Uncontrolled Resource Consumption (CWE-400) in Elasticsearch can lead to denial of service via Excessive Allocation (CAPEC-130). |
| NVIDIA Infrastructure Controller for Linux contains a vulnerability where an attacker could cause uncontrolled resource consumption. A successful exploit of this vulnerability might lead to denial of service. |
| NVIDIA Infrastructure Controller for Linux contains a vulnerability where an attacker may cause uncontrolled resource consumption. A successful exploit of this vulnerability may lead to denial of service. |
| The openml/openml.org web application version v2.0.20241110 uses predictable MD5-based tokens for critical user workflows such as signup confirmation, password resets, email confirmation resends, and email change confirmation. These tokens are generated by hashing the current timestamp formatted as "%d %H:%M:%S" without incorporating any user-specific data or cryptographic randomness. This predictability allows remote attackers to brute-force valid tokens within a small time window, enabling unauthorized account confirmation, password resets, and email change approvals, potentially leading to account takeover. |
| The brace-expansion library generates arbitrary strings containing a common prefix and suffix. Prior to 1.1.19, 2.1.5, 3.0.7, and 5.0.10, crafted brace patterns can exhaust the native stack in parseCommaParts because parseCommaParts recursively processes the remainder once per brace group and uses push.apply to pass every element of a very large comma-part array as a function argument. Patterns containing many comma-separated brace groups trigger the recursive path, while the large array triggers the argument-array path without deep recursion. These paths cause recursive and argument-array native stack exhaustion before max or maxLength can limit output, potentially terminating the Node.js process in a process-terminating denial of service. This issue is fixed in versions 1.1.19, 2.1.5, 3.0.7, and 5.0.10. |
| Net::IDN::Punycode versions from 2.302 before 2.590 for Perl leak the output buffer on every rejected label in decode_punycode.
The XS backend allocates the scalar it returns before it validates the input, sizing the buffer at twice the input length. The scalar is released only on the success path, so each of the three croaks that reject a label leaves the scalar and its buffer allocated. Nothing bounds the label length in the to-Unicode direction, since the 63-byte DNS limit is checked only when converting to ASCII.
Only the XS backend is affected.
A sender who supplies invalid labels grows the process by twice the label length per rejected call, with no successful call needed. |
| Net::IDN::Punycode versions before 2.590 for Perl allow CPU exhaustion via quadratic insertion cost when decoding a long label in decode_punycode.
The XS backend inserts each decoded code point into a UTF-8 buffer and finds the insertion point by scanning that buffer from the start, one character at a time. The scan runs once per code point over the output built so far, so the cost is quadratic in the label length. The pure-Perl backend downgrades its input to bytes so that substr can index it directly, but takes its working copy before the downgrade, so when the input carries the UTF-8 flag every substr on the copy scans from the start, with the same quadratic cost.
Nothing bounds the label length in the to-Unicode direction. The 63-byte DNS limit is checked only when converting to ASCII, so domain_to_unicode and uts46_to_unicode pass an attacker-supplied label of any length to the decoder. |
| Net::IDN::UTS46 versions before 2.590 for Perl allow CPU exhaustion via quadratic punycode encoding of an overlong label before the length check in to_ascii.
to_ascii punycode encodes each label and only then applies the 63-byte DNS limit. encode_punycode in both backends follows the sample implementation in RFC 3492, whose outer loop runs once per distinct non-ASCII code point and scans the whole input each round, so a label of distinct non-ASCII characters costs the square of its length before the limit rejects it. Every ASCII conversion in the distribution, including domain_to_ascii and email_to_ascii, goes through to_ascii. |
| In Wind River VxWorks 7 prior to 26.09, specific system call arguments can result in the process management subsystem failing to properly release allocated kernel memory before terminating the calling application. Fixed in Version 26.09 |
| Wind River VxWorks 7 24.03 through 26.03, a memory leak occurs under specific, non-default configuration states when processing specific service routines, causing the system to terminate operations before releasing allocated memory pools. Fixed in VxWorks 7 26.09 |
