| CVE |
Vendors |
Products |
Updated |
CVSS v3.1 |
| A security flaw has been discovered in Turboderp-org ExLlamaV3 up to 1.6.3. The affected element is the function ngram_dequant of the file exllamav3/exllamav3_ext/ngram.cu of the component NGramEmbedding. The manipulation of the argument head_bias results in out-of-bounds read. The attack may be launched remotely. The exploit has been released to the public and may be used for attacks. Upgrading to version 1.6.4 is sufficient to fix this issue. The patch is identified as c9a57ae832a3035e264e0a412e1daf4ebb699b98. It is suggested to upgrade the affected component. |
| Algorithm::AhoCorasick::XS versions through 0.04 for Perl read the haystack string length before the scalar is stringified.
The matches, first_match and match_details methods use the T_STD_STRING typemap to translate Perl scalars (SVs) into strings via the std::string constructor, using the SvPV macro to stringify the haystack input, and the SvCUR macro to determine the length of the SV.
When the input SVs are references, integers (IVs) or floats (NVs), the SvCUR macro will return an invalid length if it is run before the input is stringified, leading to an out-of-bounds read which can abort the process.
Note that the evaluation order of arguments to std::string is unspecified. Depending on the compiler, SvCUR may be run first and lead to an abort that cannot be caught within Perl.
This can be triggered when the haystack is a numeric value, for example,
my $ac = Algorithm::AhoCorasick::XS->new( [ "11", "22" ] );
$ac->matches( 211 );
This can occur when the haystack is the result of reading data from decoded JSON or a numeric database column. It can also be triggered when using a blessed object as a haystack. |
| FFmpeg through 9.0.2 contains a heap out-of-bounds read vulnerability in v4l2_buffer_swframe_to_buf() in libavcodec/v4l2_buffers.c that mishandles frames with negative linesize. Attackers can supply bottom-up AVI or PNM/PFM files that are transcoded to a V4L2 M2M encoder, crashing the application or leaking heap memory into encoded output. |
| FFmpeg through 9.0.2 contains an out-of-bounds read vulnerability in the SGI encoder's encode_frame() in libavcodec/sgienc.c, where unsigned size variables mis-handle negative linesize frames. Attackers can supply a video whose display matrix triggers autorotate vflip, so transcoding to SGI images reads pixel data far outside the frame and crashes the process. |
| FFmpeg before 8.1.3 and 9.x before 9.0.2 contains an out-of-bounds read vulnerability in put_vector() in libavcodec/vorbisenc.c when NaN samples leave the codebook entry at -1. Attackers can supply audio with NaN float samples to the native Vorbis encoder, causing heap reads before arrays that crash the application or leak adjacent heap memory into output. |
| FFmpeg through 9.0.2 contains an out-of-bounds read and invalid free vulnerability in aribcaption_trans_ass_subtitle() in libavcodec/libaribcaption.c that frees subtitle rectangles using region_count instead of num_rects. Attackers can supply crafted ARIB captions whose later region exceeds the 8 KiB AVBPrint limit, triggering cleanup that frees out-of-bounds pointers and corrupts the heap. |
| FFmpeg through 9.0.2 contains an out-of-bounds read vulnerability in bitpacked_decode_uyvy422() in libavcodec/bitpacked_dec.c because it never verifies packets hold width * height * 2 bytes. Attackers can supply truncated bitpacked UYVY422 packets via crafted SDP/RFC 4175 RTP sessions or raw input to crash applications or expose adjacent heap memory. |
| FFmpeg through 9.0.2 contains an out-of-bounds read vulnerability in the Cook decoder joint_decode() function because cook_decode_init() does not check js_subband_start against subbands. Attackers can supply a crafted RealMedia file whose extradata causes reads past decode_buffer_0, leaking pointers and decoder state into decoded audio to help defeat ASLR. |
| FFmpeg through 9.0.2 contains an out-of-bounds heap read vulnerability in tdsc_load_cursor() in libavcodec/tdsc.c, which stores cursor dimensions before validating them. Attackers can supply a crafted AVI file with oversized TDSC cursor dimensions so tdsc_paint_cursor() reads past the cursor buffer, crashing the decoder or leaking heap memory into frames. |
| FFmpeg through 9.0.2 contains an out-of-bounds read vulnerability in libxevd_receive_frame() in libavcodec/libxevd.c that fails to validate NAL unit length prefixes against remaining packet data. Attackers can supply crafted EVC media with oversized 4-byte NAL length prefixes, causing libxevd to read past the packet buffer, crashing the application or exposing adjacent heap memory. |
| FFmpeg through 9.0.2 contains an out-of-bounds read vulnerability in the Musepack SV7 decoder because mpc7_decode_init() does not validate lastframelen against MPC_FRAME_SIZE. Attackers can supply a crafted Musepack SV7 file declaring an oversized final frame, causing consumers of decoded frames to read past the audio buffer, leading to crashes or heap memory disclosure. |
| FastNetMon is a high-performance DDoS detector and sensor. In version 1.2.9 and earlier, process_sflow_flow_sample() in src/sflow_plugin/sflow_collector.cpp passes the attacker-controlled sflow_raw_protocol_header_t.header_size to parse_raw_packet_to_simple_packet_full() as its trusted captured length without comparing it with the bytes remaining in the record. An inflated header_size creates a false end pointer beyond the received sFlow v5 UDP buffer, defeating the nested parser's bounds checks and causing an unauthenticated remote out-of-bounds read and process crash, with possible limited exposure of adjacent bytes through parsed flow fields. No fixed release is available as of this review. |
| stable-diffusion.cpp is a pure C/C++ library for running diffusion model (Stable Diffusion, Flux, Wan, Qwen Image, Z-Image, and more) inference. Prior to master-764-e22272e, the safetensors loader in src/model_io/safetensors_io.cpp trusts each tensor's data_offsets after checking only that nbytes equals end - begin. A crafted .safetensors file can declare a byte range beyond the file's data section, causing src/model_loader.cpp to consume an attacker-chosen length from an unchecked mapped-file offset. Loading the model can read adjacent process memory into tensor buffers or crash the process. This issue is fixed in master-764-e22272e. |
| ASN1SCC is an open source ASN.1 compiler for embedded systems. Prior to 4.7.0.1, the XER tokenizer NT() in asn1crt/asn1crt_encoding_xer.c, which is reached by all XER decode paths, copies tag characters into a fixed char Value[100] field without a length bound and reads input and lookahead bytes without checking the buffer count. A crafted XER message with an overlong tag can overwrite the tokenizer's stack frame, while a truncated tag or unterminated processing-instruction or comment sequence can read beyond the input and may cause a CPU spin. The flaws can disclose adjacent memory, corrupt stack data, crash or hang the decoder, and potentially permit code execution. This issue is fixed in version 4.7.0.1. |
| Dulwich before 1.2.20 contains a commit object parsing vulnerability where the read path in _parse_commit resolves a duplicate tree header to the last occurrence instead of the first, diverging from core git's resolution behavior, while Poetry 2.5.1 ships this version and uses dulwich by default for git dependency installation. Attackers can craft a commit object with two tree headers so that reviewers and CI using git see a benign tree while Poetry's dulwich-based checkout resolves to a different attacker-controlled tree containing malicious build scripts such as setup.py, resulting in unreviewed code execution on the install or build host. |
| A flaw was found in GStreamer's MXF demuxer plugin (gst-plugins-bad). When processing a specially crafted MXF media file, the demuxer indexes its internal delta-entry and slice-offset tables using values from the file without validating that they are within the bounds of the allocated arrays. A local attacker who tricks a user into opening a malicious MXF file can cause the application to crash (denial of service) and may be able to read limited amounts of heap memory adjacent to the affected arrays. |
| unshield through 1.6.2 contains a heap out-of-bounds read vulnerability in unshield_get_file_table() that allows attackers to read past heap allocations via unchecked cabinet header offsets. Attackers can supply a crafted InstallShield cabinet with out-of-range file_table_offset or large file counts to crash the unshield tool or library consumers. |
| CGNS through 4.5.2 contains a stack out-of-bounds read vulnerability in ADFI_read_node_header that allows attackers to crash applications by supplying crafted ADF node headers. Attackers can craft .cgns or .adf files whose non-NUL-terminated end tag causes ADFI_stridx_c to scan past disk_node_data, crashing tools like cgnslist. |
| miniupnpc through 2.3.3 contains a one-byte heap out-of-bounds read vulnerability in the minixml parseelt function's namespace-prefix handling that skips an end-of-buffer check after advancing past ':'. A malicious or spoofed UPnP gateway on the local network can send a device description or SOAP reply ending in ':' within an element name to crash the client. |
| Fraunhofer HHI VVdeC before 3.2.1 contains a heap out-of-bounds read vulnerability in SEIReader::xParseSEIScalableNesting that allows attackers to read past heap allocations by casting the SEI wrapper object instead of its payload. Attackers can supply a crafted VVC bitstream with a scalable nesting SEI containing a nested buffering period SEI to crash the decoder or copy adjacent heap memory. |
