The internet is broken right now.
65 critical CVEs without a fix in the last 48 hours. Developers are crying quietly into their IDEs. 153 more already have a fix that someone still has to apply.
-
CVE-2026-125213
FFmpeg through 9.0.2 contains a heap buffer overflow vulnerability in the PSD decoder in libavcodec/psd.c that allows attackers to corrupt heap memory on 32-bit builds by supplying a crafted PSD image. Attackers can choose dimensions and channel counts whose product exceeds 4 GB, truncating the av_malloc allocation size so decode_rle() writes attacker-controlled data past the buffer, causing crashes or possible code execution.
-
CVE-2026-92454
The GUIX software JPEG decoder reads four kinds of marker segment without checking that each is long enough to hold what it reads. A truncated segment is therefore read past the end of the caller's image buffer. Separately, a zero chroma sampling factor makes the block decode loop non-terminating. Both reproduce in the default configuration, in which `GX_SOFTWARE_DECODER_SUPPORT` is enabled.
-
CVE-2026-95720
A buffer overflow vulnerability exists in AOS-8 that could allow an authenticated attacker with administrative access to cause denial of service. Successful exploitation could allow an attacker to crash the system, preventing it from rebooting without manual intervention and disrupting network operations.
-
CVE-2026-109094
A integer overflow that can lead to a buffer overrread in the XRes QueryResourceBytes Swap Path in the X.Org xwayland before 24.1.15 and X Server before 21.1.26 on 32bit systems could be used by authenticated clients to crash the X server.
-
CVE-2026-109095
A integer overflow in glamor Xv Image Size handling in X.Org xwayland before 24.1.15 and X server before 21.1.26 could be used by authenticated clients to crash the X server or leak out data.
-
CVE-2026-0454
An out-of-bounds write in the AMD Secure Encrypted Virtualization (SEV) firmware could allow a malicious hypervisor to corrupt global firmware variables, potentially resulting in loss of integrity.
-
CVE-2026-91758
Out-of-bounds Read and Write Vulnerability in Hitachi Dynamic Link Manager for Linux. This issue affects Hitachi Dynamic Link Manager for Linux: from 7.5.0-00 before 9.0.5-00.
-
CVE-2026-120434
Netpbm before 10.86.51, 11.x before 11.02.30, and 11.03 before 11.15.08 contains a stack-based buffer overflow in picttoppm's scanPoly() function in converter/ppm/picttoppm.c. Attackers can supply a crafted PICT file whose polygon edges produce more than 5000 points, overflowing the coord stack array to crash the converter or potentially execute code.
-
CVE-2026-120435
Netpbm before 10.86.51, 11.x before 11.02.30, and 11.03 before 11.15.08 contains a heap buffer overflow in xpmtoppm caused by an unvalidated characters-per-pixel value in the XPM header. Attackers can supply a crafted XPM file with characters-per-pixel greater than 3 to overrun fixed 3-byte colorName hash entries, causing crashes or potentially code execution.
-
CVE-2026-120431
Netpbm before 10.86.51, 11.x before 11.02.29, and 11.03 before 11.15.07 contains a heap-based buffer overflow in sgitopnm's readChannels() function when processing RLE-compressed SGI images. Attackers can supply crafted SGI files with oversized offset table lengths to write controlled data past the heap buffer, causing crashes or potential code execution.
-
CVE-2026-118753
FreeImage through 3.18.0 and the danoli3 FreeImage fork before 3.19.25 contain a heap buffer overflow vulnerability in PluginGIF.cpp that allows attackers to write past the canvas allocation via unclipped GIF frame rectangles. Attackers can craft a GIF whose frame left plus width exceeds the logical screen width, causing GifPlaybackCompositeFrame() and GifPlaybackFillRect() to corrupt heap memory when loaded with GIF_PLAYBACK.
-
CVE-2026-113940
An out-of-bounds read was addressed with improved bounds checking. This issue is fixed in macOS Golden Gate 27.2. An app may be able to cause unexpected system termination.
-
CVE-2026-113936
An out-of-bounds read was addressed with improved bounds checking. This issue is fixed in macOS Golden Gate 27.2. An app may be able to cause unexpected system termination.
-
CVE-2026-113938
An out-of-bounds read was addressed with improved bounds checking. This issue is fixed in macOS Golden Gate 27.2. An app may be able to cause unexpected system termination.
-
CVE-2026-111333
Memory overflow in an authentication request can lead to unpredictable or erroneous behavior or Denial of Service in Citrix NetScaler ADC and Citrix NetScaler Gateway.
-
CVE-2026-65168
NVIDIA Slurm contains an attacker-controlled heap-based buffer overflow in the slurmstepd socket handling code.
-
CVE-2026-81038
The Object Transfer Service server builds its Object Action Control Point (OACP) response indication in oacp_ind_send() (subsys/bluetooth/services/ots/ots_oacp.c). Before this fix the four-byte Checksum Value field was appended whenever the requested op code was BT_GATT_OTS_OACP_PROC_CHECKSUM_CALC, regardless of the result code. The response payload is a bare, uninitialized four-byte stack buffer declared with NET_BUF_SIMPLE_DEFINE(resp_param, sizeof(uint32_t)) in bt_gatt_ots_oacp_write(), and it is filled only on the success path of oacp_checksum_proc_validate(). On every failure path the buffer is still empty, so net_buf_simple_pull_le32() was invoked on a zero-length buffer. The OACP characteristic and its client characteristic configuration descriptor are declared with BT_GATT_PERM_WRITE and BT_GATT_PERM_READ | BT_GATT_PERM_WRITE, so no pairing, encryption, or authentication is required. Any peer within radio range that can connect to a device running the OTS server can enable OACP indications and write a nine-byte Calculate Checksum command that fails validation — for example by issuing it with no object selected, which returns BT_GATT_OTS_OACP_RES_INV_OBJ. Because oacp_command_decode() accepts the Calculate Checksum op code unconditionally (it is not guarded by CONFIG_BT_OTS_OACP_CHECKSUM_SUPPORT), builds without checksum support are reachable too: they answer BT_GATT_OTS_OACP_RES_OPCODE_NOT_SUP and take the same bad path. On builds with CONFIG_ASSERT enabled, the zero-length pull trips the assertion inside net_buf_simple_pull_mem() and panics the device, giving a remotely triggered denial of service. On builds with assertions disabled there is no check: the pull returns the untouched stack array and four bytes of uninitialized stack memory are copied into the indication and transmitted to the peer, repeatable on every request. The disclosure is bounded to four bytes per request and the write stays inside the seven-byte res indication buffer, so there is no memory corruption; the impact is uninitialized-memory disclosure plus an availability loss in assertion-enabled configurations.
-
CVE-2026-78402
Stack-based buffer overflow and out-of-bounds write vulnerability in Thales Sentinel Keys Server on Windows allows Overflow Buffers. Sentinel Keys Server does not check the length of the first field in the input message. A local attacker may craft an input and send it to Sentinel Keys Server. It may lead to code execution as NT AUTHORITY\SYSTEM enabling local privilege escalation. This issue affects Sentinel Keys Server: versions lower than 1.3.8.0.
-
CVE-2026-119372
SQLite before 3.54.0 built with FTS3/FTS4 and SQLITE_ENABLE_FTS3_PARENTHESIS contains a use-after-free vulnerability in fts3BestSnippet() caused by dangling cached position-list pointers. Attackers can run a full-text MATCH query combining OR and NEAR operators with snippet() to read freed heap memory and crash the process.
-
CVE-2026-65584
Software installed and run as a non-privileged user may conduct improper GPU system calls causing a stack overflow in the kernel. The overflow occurs when processing resolved Linux/Android fence synchronization checkpoints, while user checkpoints are excluded from the stack array. Successful exploitation can cause kernel memory corruption and typically results in a denial-of-service condition through a kernel crash. The code vulnerability is a buffer write beyond the allocated array on the stack caused by copying excessive data from an internal buffer.
-
CVE-2026-116724
FLTK through 1.4.5 contains an integer overflow vulnerability in Fl_BMP_Image::load_bmp_() that causes undersized pixel buffer allocation from file-supplied width and height. Attackers can supply a crafted BMP or ICO file with large dimensions so row decoding writes file-derived data past the heap buffer, corrupting the application heap.
-
CVE-2026-107157
Specially crafted CBOR input passed to QCborValue::fromCbor() in Qt Core can cause an out-of-bounds read, which may crash the application and could expose information.
-
CVE-2026-94192
A ThreadX module allocates the control blocks of its kernel objects from the Module Manager's object pool, and can ask for that memory back by address with `txm_module_object_deallocate`. The manager released it whatever was in it, including a control block the kernel was still using. Deallocation is not deletion: the object stays on the created list for its type, a thread stays wherever it was on the ready or suspension lists and stays schedulable, and an active timer stays on the timer list. The byte pool is meanwhile free to hand those same bytes to the next allocation.
-
CVE-2026-95109
The Cortex-A7 MMU Modules port validates a caller-supplied buffer by translating **one address** — the first byte. The common Module Manager hands its range check the module instance, the start address and the complete byte size; the Cortex-A7 implementation discarded the instance and the size and asked the MMU a single question about the start. A memory-protected module therefore passed any range whose first byte was user-readable, however far the rest of it reached. Privileged dispatch code then read or wrote the whole range. Because every protected-module dispatcher that consumes a buffer or an extra-parameter array goes through this one macro, the defect is not in any individual service: it is in the primitive all of them rely on.
-
CVE-2026-95822
The SOCKS5 client in subsys/net/lib/socks/socks.c built the CONNECT request in socks5_tcp_connect() by choosing the SOCKS5 address type from the address family of the proxy (proxy->sa_family) while reading the bytes out of the destination address supplied to connect(). A SOCKS5 request encodes the destination independently of the transport used to reach the proxy, so the two families are unrelated and nothing in the path forces them to agree: set_context_proxy() only requires the proxy family to match the socket, and zsock_connect_ctx() enters the SOCKS branch before any destination validation. With an IPv6 proxy and an AF_INET destination the code treated an 8-byte struct net_sockaddr_in as a struct net_sockaddr_in6 and copied the 16-byte sin6_addr from it, reading 12 bytes past the end of the object and transmitting them to the proxy. With CONFIG_USERSPACE enabled, z_vrfy_zsock_connect() copies only addrlen user bytes into a stack-resident struct net_sockaddr_storage and enforces no per-family minimum length, so the oversized read stays within the storage but returns stale, uninitialised kernel-stack bytes. An unprivileged user-mode thread can therefore open an AF_INET6 stream socket, point setsockopt(SO_SOCKS5) at a proxy it controls, and call connect() with a short AF_INET destination to have the kernel send it 12 bytes of supervisor stack residue per attempt. Without CONFIG_USERSPACE the read runs past the end of the caller's own struct net_sockaddr_in, disclosing adjacent memory to the proxy; the family mismatch can arise unintentionally when a destination is obtained from DNS while an IPv6 proxy is configured. Impact is limited to information disclosure of at most 12 bytes per connect() call, with no write primitive; the leak is delivered to whoever operates the SOCKS5 proxy. The mirrored pairing (IPv6 destination through an IPv4 proxy) is memory-safe but encoded the destination as SOCKS5_ATYP_IPV4 from the first four bytes of the IPv6 address, causing the proxy to connect to an unrelated host. Exploitation requires the non-default CONFIG_SOCKS option, a socket with SO_SOCKS5 configured, and both IPv4 and IPv6 support in the image. The fix moves the encoding into a helper that selects the address type from dest->sa_family and validates dest_len against the matching struct net_sockaddr_in/struct net_sockaddr_in6 before copying, rejecting under-length or unsupported destinations with an error.
-
CVE-2026-127915
The MCUboot TLV iterator bootutil_tlv_iter_next() in boot/bootutil/src/tlv.c walks image trailer TLVs with a single loop guard (it->tlv_off < it->tlv_end) and advances by sizeof(struct image_tlv) + tlv.it_len using the attacker-controlled it_len field, without validating that the entry's header or payload stays within it->tlv_end. As a result the 4-byte header read can extend up to 3 bytes past the declared TLV area, and the returned (off, len) can describe a payload that runs past it. The path is pre-authentication: TLVs are parsed during bootutil_img_validate() before the signature is verified, and a crafted image can be presented through DFU / serial-recovery / OTA. However the impact is contained by surrounding guards: bootutil_img_validate() first rejects a TLV area larger than bootutil_max_image_size(), the default LOAD_IMAGE_DATA sink is flash_area_read() which bounds-checks against the flash partition and fails an out-of-partition read, and every payload consumer (EXPECTED_HASH_TLV, EXPECTED_KEY_TLV, EXPECTED_SIG_TLV, IMAGE_TLV_SEC_CNT) clamps len against a fixed buffer before copying. The residual effect in a default flash build is therefore an out-of-region read confined to the attacker's own slot partition whose result is never observable, yielding at worst a clean validation failure rather than memory corruption or information disclosure. A genuine (though still bounded) out-of-bounds read remains under the non-default MCUBOOT_RAM_LOAD configuration, where LOAD_IMAGE_DATA is an unchecked memcpy. The fix is to validate each entry's header and payload against it->tlv_end inside the iterator.
-
CVE-2026-65138
NVIDIA Slurm contains an attacker-controlled arbitrary-memory read in the slurmstepd socket API that could be used to ex-filtrate authentication keys.
-
CVE-2026-20558
In GPU pdma, there is a possible escalation of privilege due to use after free. This could lead to local escalation of privilege with no additional execution privileges needed. User interaction is not needed for exploitation. Patch ID: ALPS11365070; Issue ID: MSV-9547.
-
CVE-2026-120754
HDF5 through 2.2.0 contains an uninitialized memory read vulnerability in H5Z__filter_scaleoffset() in src/H5Zscaleoffset.c. The decode path validates reads against the allocated size of the chunk buffer rather than the number of bytes actually read from the file. Because H5D__chunk_lock() deliberately sizes filtered chunk buffers to the larger of the stored and unfiltered chunk sizes and fills only the stored bytes, a truncated chunk passes these bounds checks and is decompressed from the uninitialized remainder of the allocation. Unrelated heap contents are then returned to the caller as dataset data. Because the reads remain within the allocation, the issue is not detected by AddressSanitizer. An attacker who can get a user or application to read a dataset from a crafted HDF5 file can cause disclosure of process memory. This issue exists because of an incomplete fix for CVE-2026-19029.
-
CVE-2026-115653
In Eclipse ThreadX versions 6.0.1 through 6.5.2.202603, the PendSV handler of the Cortex-M module ports saves a preempted user-mode module thread's registers in privileged mode at addresses taken from the thread's stack pointer, without checking that pointer against the thread's stack. A user-mode, memory-protected module can thereby write up to nine words, eight of its choosing, or 25 with floating-point state, into memory immediately below any region it may write, and read back what is stored there.
-
CVE-2026-116259
A security flaw has been discovered in libimobiledevice up to 1.4.0. This vulnerability affects the function afc_get_device_info_key of the file src/afc.c of the component Device Info Handler. Performing a manipulation of the argument Value results in memory corruption. The attack requires a local approach. The exploit has been released to the public and may be used for attacks. The vendor was contacted early about this disclosure but did not respond in any way.
-
CVE-2026-18768
A local low privileged attacker can access arbitrary physical memory or define its mapping via TcRouter.sys because of exposed IOCTL function calls resulting in a full compromise of the affected devices.
-
CVE-2026-90834
The user-mode syscall verifiers for the 16-bit ("wide data") UART API in drivers/serial/uart_handlers.c validated the wrong driver operation. z_vrfy_uart_poll_out_u16(), z_vrfy_uart_tx_u16() and z_vrfy_uart_rx_enable_u16() each ran K_SYSCALL_DRIVER_UART() against the 8-bit op (poll_out, tx, rx_enable) rather than the 16-bit op that the corresponding implementation dereferences (poll_out_u16, tx_u16, rx_enable_u16). z_impl_uart_poll_out_u16(), z_impl_uart_tx_u16() and z_impl_uart_rx_enable_u16() in include/zephyr/drivers/uart/uart_internal.h call their op unconditionally, with no NULL guard of their own, because the verifier is supposed to have established that the op exists. CONFIG_UART_WIDE_DATA is an application-selectable option that no driver selects, and only a handful of the serial drivers in the tree implement any _u16 operation; every other driver leaves those struct uart_driver_api members NULL while still providing the 8-bit ops the old check inspected. In a build with CONFIG_USERSPACE and CONFIG_UART_WIDE_DATA (plus CONFIG_UART_ASYNC_API for the uart_tx_u16() and uart_rx_enable_u16() variants), an unprivileged user-mode thread that legitimately holds access to such a UART device object can invoke the _u16 syscall: the object, permission and op checks all pass, and the kernel-side implementation then performs an indirect call through a NULL function pointer while running at supervisor privilege. The result is a CPU exception taken in kernel context instead of the intended K_OOPS or -ENOSYS return. With the default k_sys_fatal_error_handler() this halts the system (or warm-reboots it under CONFIG_RESET_ON_FATAL_ERROR), so a thread confined by the userspace sandbox can deny service to the whole device — a capability the user-mode privilege boundary is specifically meant to withhold. The call target is a fixed NULL, so there is no attacker-controlled code execution and no memory disclosure; the separate K_SYSCALL_MEMORY_ARRAY_READ()/K_SYSCALL_MEMORY_ARRAY_WRITE() bounds checks on the caller's buffer were and remain correct. The fix points each verifier at the matching _u16 op, using a plain K_SYSCALL_OBJ() object check where the op does not exist or where the implementation already tests it for NULL.
-
CVE-2026-125262
FFmpeg through 9.0.2 contains a heap buffer overflow vulnerability in libavcodec/lcevcdec.c because LCEVC output picture dimensions are not validated against the allocated frame. Attackers can embed crafted LCEVC enhancement data with differing global configurations in H.264 or HEVC streams to corrupt heap memory in decoding applications.
-
CVE-2026-125250
FFmpeg through 9.0.2 contains a heap buffer overflow in the GIF encoder's gif_image_write_image() in libavcodec/gif.c because shrunk_buf is sized from the first frame's linesize. Attackers can supply PAL8 video whose frame linesize grows mid-stream, causing remap_frame_to_palette() to write past shrunk_buf and corrupt the heap during GIF transcoding.
-
CVE-2026-123203
NEURON (CoreNEURON) through 9.0.2 contains a stack buffer overflow vulnerability that allows attackers to corrupt stack memory by supplying over-long tokens in a crafted model dataset. Attackers can place oversized version or mechanism-name tokens in bbcore_mech.dat, globals.dat, or files.dat loaded by nrniv-core to crash the process.
-
CVE-2026-68993
A double free vulnerability in NPU driver due to improper pointer management (940)
-
CVE-2026-121511
Heap Out-of-Bounds Read in µD3TN BPv6/BPv7 Fragmenter allows Local Attacker to Read Memory and Reliably Cause DoS
-
CVE-2026-78119
The Nuvoton NPCX eSPI driver parses Target-Attached-Flash (TAF) requests in espi_taf_parse() (drivers/espi/espi_npcx.c), taking the 12-bit length field directly out of the request header in the FLASHRXBUF FIFO. For write and RPMC OP1 requests it copied DIV_ROUND_UP(len, 4) words into struct espi_taf_pckt.src, a fixed 16-word (64-byte) member of a static structure, without ever comparing the length against that buffer or against the negotiated eSPI receive payload size (rx-plsize, 64 bytes on all in-tree NPCX/NPCK SoCs). Both the loop counter and the word count were uint8_t, so the copy ran up to 255 iterations. An eSPI controller that emits a flash request whose header claims a payload larger than the negotiated maximum therefore drives up to 239 32-bit stores past the end of the structure — roughly 956 bytes of out-of-bounds write into .bss — from the eSPI flash receive interrupt handler, while the source reads walk past the end of the 18-word FLASHRXBUF register array into neighbouring eSPI registers. The same unvalidated length was then passed on to flash_write() in espi_taf_npcx_flash_write() with the 64-byte src buffer as source, reading up to 4 KB of adjacent controller SRAM into SPI flash at a host-chosen address, where the host can read it back. A related fixed-size stack buffer in taf_npcx_completion_handler() was overflowed by four bytes on the maximum-size completion path on SoCs whose tx-plsize is 64. The defect is reachable only from the eSPI host side, and only by a request that violates the negotiated payload size, so it requires host firmware or bus-level control rather than ordinary host software; a compliant maximum-length write fits the buffer exactly and is handled correctly. Because TAF exists specifically so that the embedded controller arbitrates host access to shared flash (espi_taf_check_read_protect() / espi_taf_check_write_protect()), corrupting controller memory from a host-supplied field crosses the trust boundary the feature asserts; consequences range from controller crash or hang to corruption of driver state that gates the flash protection regions. The fix widens the counters, rejects any request whose word count exceeds both ARRAY_SIZE(FLASHRXBUF) - NPCX_ESPI_FLASH_HEADER_ADDRESS_LEN and NPCX_TAF_RX_PAYLOAD_WORDS (derived from the devicetree rx-plsize), marks such packets with a new invalid_flag so espi_taf_work() answers them with an unsuccessful completion instead of processing them, bounds the transmit payload against FLASHTXBUF, and adds a default case for unsupported cycle types. Only configurations with CONFIG_ESPI_TAF enabled on an NPCX/NPCK target are affected; no in-tree board enables it by default.
-
CVE-2026-107493
Issue summary: When an application passes a stack to PEM_X509_INFO_read_bio_ex() to add entries to, and reading fails, OpenSSL frees every entry in the stack, both the old ones and any new ones, but leaves them in the stack. Impact summary: An application that passes an existing stack to PEM_X509_INFO_read_bio_ex() and keeps using it after a failed read may use memory that was already freed. This can cause a crash or, in principle, memory corruption. Malformed PEM data can make the read fail, but only applications that pass an existing stack are affected, and OpenSSL itself does not do this. CWE: CWE-416: Use After Free Description: PEM_X509_INFO_read_bio_ex() reads certificates, CRLs and private keys from PEM data and adds them to a stack of X509_INFO structures. If the application passes a stack that is not NULL, the function adds the new entries to it and must leave the entries that are already there unchanged. If reading fails, the function frees every entry in the stack. This includes the entries that were there before the call, and the entries added by the call. It does not remove these entries from the stack, so the stack keeps pointers to freed memory. The failure can happen at any point, even before any entry was added. A PEM item with invalid content is enough. This also happens if the stack was empty: entries added before the failure are freed but stay in the stack. An application that then reads or frees the stack will use memory that was already freed. This can crash the application or, in principle, corrupt memory. Only applications that pass a stack that is not NULL are affected. OpenSSL itself never does this. Its own callers, including the code that TLS uses to load CA certificates, always pass NULL. When the stack is NULL, the function frees only the entries that it created itself, which is safe. We expect that very few applications pass an existing stack. An attacker can make the read fail with malformed PEM data, but cannot make an application pass an existing stack. The same problem affects PEM_X509_INFO_read_bio(), PEM_X509_INFO_read() and PEM_X509_INFO_read_ex(), which use the same code. OpenSSL 1.1.1 and 1.0.2 have only the first two of these functions. FIPS impact: no The affected code is outside the FIPS module boundary.
-
CVE-2026-125260
FFmpeg through 9.0.2 contains a heap buffer overflow vulnerability in libavcodec/exrenc.c that allows attackers to corrupt heap memory when encode_scanline_rle() processes frames taller than the encoder height. Attackers can supply crafted input whose resolution grows mid-stream, transcoded to OpenEXR with RLE compression, to write pointers past the scanline array and crash the application.
-
CVE-2026-125256
FFmpeg through 9.0.2 contains a heap buffer overflow vulnerability in ff_llvidenc_diff_bytes_rvv() that allows attackers to corrupt heap memory by triggering negative length values in the Huffyuv encoder. Attackers can supply RGB video narrower than 8 or 16 pixels for huffyuv transcoding on RVV-enabled RISC-V builds, causing writes past the encoder's temp buffer.
-
CVE-2026-125257
FFmpeg through 9.0.2 contains an off-by-one heap overflow vulnerability in get_exponent_dynamic() in libavcodec/nellymoserenc.c that allows writing one element past the opt and path allocations. Attackers can supply crafted audio to applications encoding Nellymoser with trellis enabled, causing idx to equal OPT_SIZE and overwrite adjacent heap memory.
-
CVE-2026-125227
FFmpeg through 9.0.2 contains an integer overflow in cbs_jpeg_write_scan() in libavcodec/cbs_jpeg.c that causes a heap buffer overflow on 32-bit builds. Attackers can supply a crafted MJPEG/JPEG stream with a scan of 512 MiB or more, processed by CBS bitstream filters like filter_units, to corrupt the heap or crash the application.
-
CVE-2026-125218
FFmpeg through 9.0.2 contains a heap buffer overflow vulnerability in the APNG encoder's encode_apng() function in libavcodec/pngenc.c when frame size changes mid-stream. Attackers can supply crafted media whose resolution grows mid-stream, causing APNG_DISPOSE_OP_BACKGROUND memset writes past prev_frame and crashing or corrupting heap memory in transcoding applications.
-
CVE-2026-125219
FFmpeg through 9.0.2 contains a heap buffer overflow vulnerability in vdpau_hevc_start_frame() in libavcodec/vdpau_hevc.c that copies unbounded PPS tile counts into fixed-size VDPAU arrays. Attackers can supply crafted HEVC bitstreams with excessive tile columns or rows to write controlled 16-bit values past heap-allocated hwaccel data, corrupting memory in applications using VDPAU decoding.
-
CVE-2026-125212
FFmpeg through 9.0.2 contains a heap buffer overflow vulnerability in libavcodec/videotoolboxenc.c that copies frames larger than encoder dimensions into undersized CVPixelBuffers. Attackers can supply crafted input whose resolution grows mid-stream, transcoded with VideoToolbox encoders without rescaling, to crash the application or corrupt memory.
-
CVE-2026-125214
FFmpeg through 9.0.2 contains a heap buffer overflow vulnerability in pcm_bluray_decode_frame() in libavcodec/pcm-bluray.c on big-endian hosts. Attackers can supply a crafted Blu-ray (M2TS) LPCM stream whose payload length is not a multiple of the sample size, writing attacker-controlled bytes past the audio frame buffer.
-
CVE-2026-125210
FFmpeg through 9.0.2 contains a heap buffer overflow vulnerability in aribcaption_trans_text_subtitle() in libavcodec/libaribcaption.c when decoding with sub_type text. Attackers can supply a crafted ARIB STD-B24 caption stream with zero regions, causing a pointer write past a zero-sized sub->rects allocation and corrupting heap memory.
-
CVE-2026-125202
FFmpeg through 9.0.2 contains a heap buffer overflow vulnerability in vaapi_hevc_start_frame() in libavcodec/vaapi_hevc.c that copies unbounded PPS tile counts into fixed-size arrays. Attackers can supply a crafted HEVC bitstream with a large picture and many tiles to write controlled 16-bit values past heap-allocated hwaccel data, corrupting memory in applications decoding with VAAPI.