| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| SIPGO is a library for writing SIP services in the GO language. Prior to 1.4.1, ParserStream.parseSingle in sip/parser_stream.go allocates a SIP body buffer from the client-controlled Content-Length header before ParseMaxMessageLength is enforced. An unauthenticated peer can send a stream-transport message over TCP, TLS, WS, or WSS with an oversized declared length, causing excessive memory allocation and denial of service before the body is read. This issue is fixed in version 1.4.1. |
| lwIP TCP/IP Stack MQTT is vulnerable to an out-of-bounds write, which may allow an attacker to gain full code execution on the device. |
| psd-tools is a Python package for working with Adobe Photoshop PSD files. Prior to 1.17.4, PSDImage.composite() and PSDImage.numpy() allocated output buffers from attacker-controlled PSD header geometry, including width, height, channels, depth, and per-layer rectangles, before validating those values against the available file data. A tiny crafted PSD could therefore cause multi-gigabyte memory allocation, and PSDImage.composite() could return a black image with only a warning instead of raising an exception. Services that composite untrusted PSD files could be terminated by out-of-memory handling. This issue is fixed in version 1.17.4. |
| vLLM through 0.29.0 fails to validate the tp_size parameter in kv_transfer_params on OpenAI-compatible completion endpoints, allowing attackers to allocate unbounded memory. Attackers can supply arbitrary tp_size values in prefill/decode disaggregated deployments to exhaust memory and trigger kernel OOM-kill of the decode worker process. |
| vLLM through 0.29.0 contains a memory corruption vulnerability in the Triton _bincount_kernel where prompt token IDs index the penalty prompt-presence bitset without bounds checking against vocabulary size. Attackers can submit multimodal audio requests with tokens equal to vocabulary size, causing out-of-bounds writes that corrupt concurrent requests' sampler state and alter repetition penalty behavior. |
| vLLM through 0.29.0 fails to properly validate bad_words token indices against the model's generation output width in SamplingParams.update_from_tokenizer(). Attackers can supply out-of-bounds token indices that corrupt logits memory of concurrent requests, causing different in-flight HTTP requests to return incorrect tokens. |
| snappy-java through 1.1.10.8 contains a buffer overflow vulnerability in typed Snappy.uncompress*Array methods that allocate output arrays by dividing uncompressed length by element size but pass the undivided length to native code. Attackers controlling compressed input can cause misaligned length values to write past array bounds with attacker-controlled bytes, corrupting heap memory. |
| vLLM before 0.29.0 validates allowed_token_ids against tokenizer length instead of model output logits width in SamplingParams._validate_allowed_token_ids(). Attackers can supply token IDs above the output vocabulary that pass validation, causing LogitBiasState to corrupt GPU logits state and allow concurrent requests to sample tokens outside their allowlists. |
| Email::Sender::Transport::Sendmail versions before 2.602 for Perl allow arbitrary command execution on Windows sending a message whose envelope address reaches the shell in _sendmail_pipe.
On MSWin32 the envelope sender and every recipient go into a single command string, which open() passes to a shell. Every other platform gets the list form, which runs sendmail directly. When the caller supplies no envelope, Email::Sender::Simple takes the recipients from the To and Cc headers and the sender from the From header.
An attacker who controls one of those header addresses runs commands as the sending process. |
| A security flaw has been discovered in Moonshot AI Kimi Code up to 0.31.0. The affected element is an unknown function of the file agent-core-v2/src/agent/mcp/config-loader.ts of the component MCP Configuration Loader. The manipulation results in os command injection. The attack may be launched remotely. The exploit has been released to the public and may be used for attacks. Upgrading to version 0.31.1 is sufficient to fix this issue. It is recommended to upgrade the affected component. Beyond the trust prompt, the fix resolves fd/stty binaries to absolute paths specifically "so untrusted workspaces cannot plant bare-name executables before confirmation," fixing a secondary $PATH path-planting vector alongside the primary untrusted-.mcp.json auto-spawn. |
| The Okta Access Gateway does not neutralize shell metacharacters in SNMP configuration values before a privileged script uses them to construct OS commands. An authenticated local user with access to the management interface can supply crafted values, resulting in the execution of arbitrary OS commands with root privileges. |
| Wazuh is a free and open source platform used for threat prevention, detection, and response. From 3.9.0 until 4.14.5 and 5.0.0-beta2, the Wazuh cluster protocol in framework/wazuh/core/cluster/common.py allows an authenticated cluster node to exhaust memory on the master. The receive_str() method accepts an attacker-controlled total for InBuffer without a maximum, so a new_str command can request a multi-gigabyte bytearray and repeated requests accumulate in in_str. The divided-message path also retains flag_divided fragments under unique counters in div_msg_box without a count, aggregate-size, or expiration limit. Exploitation can disrupt agent connectivity and alert processing across the monitored environment. This issue is fixed in versions 4.14.5 and 5.0.0-beta2. |
| FreeRDP is a free implementation of the Remote Desktop Protocol. Prior to 3.27.0, FreeRDP clients that negotiate RDPGFX AVC444 with an H.264 decoder backend calculate the intermediate YUV444 allocation size in libfreerdp/codec/h264.c with 32-bit multiplication in avc444_ensure_buffer. A malicious RDP server can supply surface dimensions for which piDstStride multiplied by padDstHeight wraps to a small nonzero value, causing winpr_aligned_recalloc to allocate an undersized buffer before YUV420CombineToYUV444 writes using the actual stride and rectangle dimensions. This can cause a client crash and may permit code execution through attacker-influenced heap corruption. This issue is fixed in version 3.27.0. |
| A malicious X server could exploit a buffer overflow in libX11 before 1.8.14 during handling of XkbGetMap overflowing the key_sym_map. |
| A heap-based buffer overflow was found in the DHCPv6 and TFTP response builders of libslirp. When the host is configured with a small interface MTU, a guest-supplied DHCPv6 CLIENTID option or TFTP blksize option can overflow the reply buffer with attacker-controlled content and length, resulting in denial of service and potentially arbitrary code execution in the host process. The default interface MTU is not affected. |
| Memory corruption when processing decode statistics due to insufficient validation of offset against structure size. |
| Memory Corruption when processing data with large offset and length values exceeds buffer limits during data copy operations. |
| MaxKB is an open-source AI assistant for enterprise. Prior to 2.10.5-lts, authenticated workspace members can inject control characters into AWS Bedrock access_key_id and secret_access_key fields that _update_aws_credentials writes to /root/.aws/credentials without safe parsing. An attacker can append a new AWS profile containing credential_process, then select that profile during a later model-validation request so botocore executes an attacker-controlled command as root. This vulnerability is fixed in 2.10.5-lts. |
| MaxKB is an open-source AI assistant for enterprise. Prior to version 2.10.5-lts, assistants with a tool, MCP tool, skill, or sub-application use SandboxShellBackend, which exposes an execute shell tool without excluding it and omits execute from interrupt_on, so human approval is not required. Untrusted chat or ingested content can therefore cause command execution; source deployments with MAXKB_SANDBOX disabled run commands directly as the application user, while the official root container's string-based gosu wrapper allowed shell metacharacters to execute outside the intended sandbox. This issue is fixed in version 2.10.5-lts. |
| Imager::File::PNG versions from 1.003 before 1.004 for Perl write past the end of the row buffer reading a PNG with a tRNS transparency chunk in read_direct8.
With a tRNS chunk, read_direct8() adds an alpha channel to the image it creates but still sizes the row buffer from the original channel count. libpng expands the transparency into that extra channel, so png_read_row() fills one channel more than the buffer holds, at one byte per sample, and writes width bytes past the end of the allocation. Palette images go to read_paletted() and 16-bit images to read_direct16(), which sizes its buffer from png_get_rowbytes() and allocates enough for the expanded row.
The same reader ships bundled in the Imager distribution.
Reading an attacker-supplied PNG through Imager->read() corrupts the heap, which can crash the process. |
