| 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. |
| Allocation of Resources Without Limits or Throttling vulnerability in elixir-mint mint allows a malicious HTTP/2 server to make the client hold up to about 16 MiB per connection in frames it should reject, consuming client memory.
Mint.HTTP2.Frame.decode_next/2 in lib/mint/http2/frame.ex compares a frame with the client's max_frame_size (16,384 bytes by default) only once the whole declared payload has arrived. Until then it returns :more, and Mint.HTTP2 keeps every received byte in the connection buffer. A server can declare a frame length of up to 16,777,215 bytes and withhold the last byte, keeping roughly 1,024 times the advertised limit buffered for as long as the connection stays open. The server has to send every byte the client buffers, so there is no amplification, and the buffer stops at the 24-bit frame length limit.
This issue affects mint: from 0.1.0 before 1.10.2. |
| Unauthenticated Denial of Service Attack in WP Store Locator < 3.0.0 versions. |
| Snipe-IT is an IT asset/license management system. Prior to 8.6.1, POST /two-factor has no rate limiting, lockout, or attempt counter, allowing an attacker with valid credentials to submit unlimited TOTP guesses against the three accepted codes created by config/google2fa.php window=1. A successful guess creates a fully authenticated session. When two_factor_enabled is 1, POST /account/profile with two_factor_optin=0 can disable two-factor authentication without OTP reverification, while required mode 2 prevents that opt-out. An administrator can also use POST /api/v1/users/two_factor_reset to clear another user's secret. This issue is fixed in version 8.6.1. |
| Unauthenticated Denial of Service Attack in Two Factor <= 0.16.0 versions. |
| crmne/ruby_llm at commit fa6f279847d6d7027814539d9c0dfc3bbdfd2a83 contains polynomial-time regular expression denial-of-service conditions in think-tag response parsing on Ruby 3.1.x. A malicious or anomalous model response containing many unterminated <think> tags can cause excessive CPU consumption in two consecutive regular expressions and delay chat-completion processing |
| In the Linux kernel, the following vulnerability has been resolved:
ksmbd: rate limit unmapped SID errors
A client can include many structurally valid but unmapped SIDs in a DACL.
Logging every mapping failure lets one request generate hundreds of kernel
error messages.
Rate limit the message to prevent an authenticated client from flooding
the kernel log. |
| Apache XmlSchema doesn't limit how deeply schema structures can be nested when it builds its schema model, so a malicious schema can make parsing recurse until the stack overflows. This causes a denial of service.
Users are recommended to upgrade to version 2.3.3, which fixes this issue. |
| LightLLM through 1.2.0 contains a memory exhaustion vulnerability in the NCCL control channel when started with --pd_trans_mode nccl, allowing unauthenticated attackers to exhaust KV-transfer worker memory. Attackers can call the exposed_set_value method to store unbounded key-value pairs without size limits, causing the worker process to crash and triggering node failure. |
| The TFTP server accepts a DATA datagram of any size. The dispatcher rejects datagrams shorter than
four bytes (nxd_tftp_server.c:1037) and nothing anywhere checks an upper bound, in particular not
against the protocol maximum of 4 + NX_TFTP_FILE_TRANSFER_MAX. Two things follow from that one
missing check, both reachable before any authentication because TFTP has none.
The handler passes `nx_packet_length - 4` straight to FileX:
```c
/* addons/tftp/nxd_tftp_server.c:1863, 1889 */
status = nx_packet_copy(packet_ptr, &temp_ptr,
server_ptr -> nx_tftp_server_packet_pool_ptr, NX_WAIT_FOREVER);
...
fx_file_write(&(client_request_ptr -> nx_tftp_client_request_file),
packet_ptr -> nx_packet_prepend_ptr + 4,
packet_ptr -> nx_packet_length - 4);
```
`nx_packet_length` is the length of a chain, not of one contiguous buffer, so FileX copies past the
end of the first packet:
```
ERROR: AddressSanitizer: heap-buffer-overflow
READ of size 1280 at 0x621000001108 thread T5
#0 __interceptor_memcpy
#1 _fx_utility_memory_copy filex/common/src/fx_utility_memory_copy.c:78
0x621000001108 is 0 bytes to the right of 4104-byte region
```
Those bytes are written into the file the attacker is uploading, and a TFTP read request hands them
back, so this is a memory disclosure with a convenient retrieval channel.
The same datagram also wedges the server. `nx_packet_copy` at :1863 needs
ceil(nx_packet_length / pool_payload) packets and asks for them with NX_WAIT_FOREVER, so when the
attacker sizes the datagram beyond what the pool holds, the server thread suspends and never
returns. A liveness probe after one such datagram times out with the pool at 0 of 12 packets and
the server thread suspended, and no later client is served.
Reject `nx_packet_length > 4 + NX_TFTP_FILE_TRANSFER_MAX` in the DATA branch before either call,
and use a bounded wait rather than NX_WAIT_FOREVER for the copy. |
| Issue summary: A malicious remote peer may flood the local QUIC
stack with NEW_CONNECTION_ID frames by avoiding a limit check on
how many connection IDs the remote QUIC stack can use.
Impact summary: The local QUIC stack sends a RETIRE_CONN_ID frame
for every NEW_CONNECTION_ID frame it receives. The RETIRE_CONN_ID
frame is dispatched via the Control Frame Queue (CFQ). If the remote
peer also withholds ACKs, then it can force the local stack
to allocate ~400MB (depending on ACK delay).
CWE: CWE-770: Allocation of Resources Without Limits or Throttling
Description: RFC 9000 sections 5.1.1 and 5.1.2 [1] describe the mechanism
by which a remote peer can notify the local QUIC stack to change the
destination connection ID (a.k.a. CID) the local stack uses to
identify the connection at the remote peer. Each CID is associated
with a sequence number. The sequence number is transmitted
in NEW_CONNECTION_ID and RETIRE_CONNECTION_ID frames to identify the CID
which is being either associated with a connection or retired.
The remote peer sends a NEW_CONNECTION_ID frame to let the local stack know
a new CID is being associated with an existing connection. The
NEW_CONNECTION_ID frame carries the new CID, its sequence number, and the
retire-prior-to number. The retire-prior-to identifies existing
CIDs that are to be retired. The local QUIC stack must send a
RETIRE_CONNECTION_ID for every destination CID whose sequence number
is less than retire-prior-to. The CID becomes retired after the
local stack receives an ACK for its RETIRE_CONNECTION_ID frame.
Although the OpenSSL QUIC stack supports at most one destination CID
for every connection, it can be tricked into processing more than
one RETIRE_CONNECTION_ID frame per connection. The OpenSSL QUIC
stack currently retires the destination CID as soon as it receives
the NEW_CONNECTION_ID, while in fact the destination CID must
be retired after an ACK for the RETIRE_CONNECTION_ID frame is received.
Correcting the flawed logic also fixes the backlog growth.
[1] https://datatracker.ietf.org/doc/html/rfc9000#name-issuing-connection-ids
FIPS impact: no
The FIPS module is not affected as the QUIC implementation is outside of
the OpenSSL FIPS module boundary. |
| Issue summary: QUIC process may keep memory for QUIC packet
buffer for much longer period than necessary.
Impact summary: Remote peer can exploit this vulnerability
by sending maliciously crafted packets, making the local
QUIC stack to keep the memory for packet buffers allocated.
The time for which the memory remains allocated is entirely
under the control of the potentially malicious remote peer.
CWE: CWE-770: Allocation of Resources Without Limits or Throttling
Description: To save copy operation from the packet buffer to the
stream reassemble buffer the QUIC stack leaves the stream data
on the packet buffer waiting to be copied to a buffer provided
by the local receiving application. The QUIC stack releases
a reference to the packet buffer only after the data are copied
to the application buffer. This design is more efficient for
legitimate data transfers but enables an attacker to allocate a lot
more memory than actually required by the data kept in the receiving
stream buffer.
To mitigate the vulnerability, the QUIC stack now calculates
and monitors memory overhead for every stream. The memory overhead
for a single stream frame is calculated as a difference between the
size of the whole packet that carries the stream frame and the size
of the stream frame itself. The memory overhead for a single stream
frame is added to the total (cumulative) memory overhead QUIC stack
keeps for each stream. Once the cumulative memory overhead exceeds
64kB, the QUIC stack moves the stream frame data from the packet
buffer to the stream buffer, starting with the next packet received.
FIPS impact: no
The FIPS module is not affected as the QUIC implementation is outside of
the OpenSSL FIPS module boundary. |
| Issue summary: The QUIC stream reassembly algorithm performance deteriorates
progressively as packets are arriving out of order. The worst case has
a quadratic complexity proportional to the number of stream frames kept in
the buffer for the received stream data.
Impact summary: A remote QUIC peer that completes the handshake can create
a connection-scoped CPU pressure and potentially a Denial of Service using
compliant STREAM frames inside the advertised receive window, with low
attacker bandwidth.
CWE: CWE-407: Inefficient Algorithmic Complexity
Description: OpenSSL manages received QUIC stream fragments using a
doubly-linked list. While it optimizes for append operations (at the end of
the list), it falls back to a head-to-tail linear search for any fragment
that does not immediately follow the current `tail`.
By manipulating the sequence of offsets, an attacker can force the server
to perform O(n^2) operations, consuming excessive CPU time for the
QUIC process.
FIPS impact: no
The FIPS module is not affected as the QUIC implementation is outside of
the OpenSSL FIPS module boundary. |
| Issue summary: A certificate with many nameRelativeToCRLIssuer CRL
distribution points causes disproportionate heap growth when OpenSSL caches
X.509 extensions.
Impact summary: Receiving a crafted certificate from a malicious peer can lead
to significant memory pressure and possible Denial of Service in clients or
in servers that solicit client certificates.
CWE: CWE-770: Allocation of Resources Without Limits or Throttling
Description: A certificate or a set of certificates that fits under the limit for
size of certificates accepted from the peer (~100 KiB) can result in allocation
of several hundred MiB of resident memory on the receiving side
during a normal TLS handshake. This may be enough to crash the client or
server, if multiple concurrent connections lead to similarly large memory
allocations.
The fix postpones processing of the CRL distribution points extensions in
certificates to the time when the processed value is required for CRL processing.
This avoids keeping large memory allocations for a long time when such
certificates are received.
FIPS impact: no
The affected code is outside the FIPS module boundary. |
| Issue summary: OpenSSL QUIC stack does not enforce connection
level flow control for streams. Remote peers may send more bytes
as long as they fit within the stream flow control limits.
Impact summary: A malicious remote peer may exploit the lack of connection
flow control for streams to make the QUIC stack receive ~100MB of memory
instead of 768 KiB (default flow control window size).
CWE: CWE-770: Allocation of Resources Without Limits or Throttling
Description: The local QUIC stack advertises two flow control limits
to its remote peer: stream flow control limit and connection flow
control limit. The remote peer must follow both limits when transmitting
stream data.
Whenever the local QUIC stack receives a stream frame, it validates
that the size of the received stream frame stays within flow control limits.
If either limit is exceeded (stream level or connection level), then
the QUIC stack must close the connection with a flow control error.
The vulnerable OpenSSL QUIC stack enforces the stream-level but not
the connection-level limit. To exploit the issue, three conditions must be met:
- the remote peer opens several streams
- each stream must stay within the stream-level flow control limit
- there must be no zero-offset byte sent on any of the streams
(to prevent the vulnerable QUIC stack from consuming data).
By meeting the conditions above, the remote peer may make the local stack
allocate 2 x MAX_STREAMS x (stream flow control limit) bytes
of memory. MAX_STREAMS defaults to 100, and the limit applies to both
bidirectional and unidirectional streams, making it 200 in total. The default
flow control window for a stream is 512kB. The remote peer may
force the vulnerable QUIC stack to allocate 100MB of heap per connection.
FIPS impact: no
The FIPS module is not affected as the QUIC implementation is outside of
the OpenSSL FIPS module boundary. |
| Apache XmlSchema doesn't limit how deeply schema imports and includes can be nested, so a malicious schema can make parsing recurse until the stack overflows. This causes a denial of service.
Users are recommended to upgrade to version 2.3.3, which fixes this issue. |
| In the Linux kernel, the following vulnerability has been resolved:
smb/client: validate new EOF for zero range
When FALLOC_FL_ZERO_RANGE is used without FALLOC_FL_KEEP_SIZE,
smb3_zero_range() may extend EOF without checking RLIMIT_FSIZE, allowing
the file to grow beyond the caller's file-size limit.
Fix this by calling inode_newsize_ok() before sending the zero-range
request when the operation would extend EOF.
Reproducer, using a file on a CIFS mount:
bash -c '
FILE=/mnt/cifs/repro
trap "" SIGXFSZ
ulimit -f 3072
truncate -s 2M "$FILE"
fallocate --zero-range -o 0 -l 4M "$FILE"
echo "fallocate rc=$?"
stat -c "file size=%s" "$FILE"
'
Before this change, the operation succeeds despite the 3 MiB limit:
fallocate rc=0
file size=4194304
After this change, fallocate fails and leaves the file at 2 MiB. |
| 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) |
| RabbitMQ is a messaging and streaming broker. From 3.13.0 until 3.13.15, 4.0.20, 4.1.11, and 4.2.6, the Web STOMP WebSocket handler enforced neither max_frame_size nor login_timeout before authentication, allowing an unauthenticated client to keep a connection alive with a slow stream of small frames and accumulate unbounded pre-authentication state. The rabbitmq_web_stomp plugin must be enabled, and no authentication is required to reach the vulnerable path. This issue is fixed in versions 3.13.15, 4.0.20, 4.1.11, and 4.2.6. |
