Pillow is a Python imaging library. Prior to 13.0.0, Pillow's JPEG decoder can write a full-width scanline into a buffer sized for a requested draft scale. ImagingJpegDecode does not verify libjpeg's actual output width and component count after scaling is requested, and an 8-bit lossless SOF3 JPEG decoded by a build linked to libjpeg-turbo 3.x can ignore the requested downscaling geometry. An attacker can supply such a JPEG to an application that uses draft() or thumbnail() to request downscaling. This can corrupt heap memory, crash the process, or potentially allow code execution. Older libjpeg-turbo 2.x and IJG builds reject the SOF3 input. This issue is fixed in version 13.0.0.
Pillow is a Python imaging library. Prior to 13.0.0, Pillow can double free image memory when ImagingScaleAffine encounters a selective allocation failure. The affected fast path handles Transform.EXTENT or pure scale-and-translate Transform.AFFINE operations with the default NEAREST resampling. The output image allocation must succeed while the additional per-column lookup allocation fails, after which the error path deletes image storage that later cleanup releases again. An attacker who can drive such a large Image.transform() operation under memory pressure may trigger the failure. This can crash the process and may corrupt memory. This issue is fixed in version 13.0.0.
Pillow is a Python imaging library. From 5.4.0 until 13.0.0, Pillow performs quadratic copying while parsing GIF comments. GifImageFile._seek() repeatedly appends immutable bytes for comment sub-blocks and multiple comment extensions, and frame zero is parsed during Image.open(). An attacker can submit a small-dimension GIF containing many comment extensions to any application that opens the file. This can consume excessive CPU and deny service before the application can inspect the image. This issue is fixed in version 13.0.0.
Pillow is a Python imaging library. From 9.2.0 until 13.0.0, Pillow can retain a rejected GIF frame size after a decompression-bomb exception. In GifImageFile._seek, a crafted GIF can leave rejected dimensions stored after Image.DecompressionBombError; if an application catches that exception and continues processing the same image object, later seek or load operations allocate according to those dimensions and can exhaust memory, denying service. This issue is fixed in version 13.0.0.
Pillow is a Python imaging library. From 8.1.0 until 13.0.0, Pillow can allocate an APNG disposal background at the full canvas size before applying the decompression-bomb check. A crafted APNG with a small first frame, a very large declared canvas, and background disposal or first-frame OP_PREVIOUS reaches PngImageFile._seek while Image.open() positions frame zero. The full-canvas allocation occurs inside Image.open() without requiring pixel loading or frame iteration. This can exhaust memory and deny service. This issue is fixed in version 13.0.0.
Pillow is a Python imaging library. Prior to 13.0.0, Pillow's FlashPix JPEG-tile decoder can write decoded scanlines beyond the allocated tile buffer. FlashPix (FPX) JPEG decoding selects raw mode LA with two bytes per pixel, while an unrecognized JPEG mode can be decoded as JCS_UNKNOWN with four components, causing ImagingJpegDecode to write four-component scanlines into the two-byte tile buffer. An attacker can supply a crafted OLE2 FlashPix file to an application with the optional olefile dependency installed that opens or thumbnails untrusted input. This can corrupt heap memory and may allow arbitrary code execution. This issue is fixed in version 13.0.0.
Pillow is a Python imaging library. Prior to 13.0.0, Pillow's ImageMorph.MorphOp.apply() can access memory before a row buffer for a zero-width image. PPM and PGM input can create an image with width zero, and the native morphology routine indexes outrow[width - 1] without rejecting that geometry. An attacker can supply such an image to an application that applies a morphology operation. This can crash the process. This issue is fixed in version 13.0.0.
Pillow is a Python imaging library. From 11.3.0 until 13.0.0, Pillow has a heap use-after-free when libtiff rejects a tag while encoding TIFF data to an in-memory buffer. When an application saves attacker-influenced TIFF tags to BytesIO, libtiff can reject a tag during encoding in ImagingLibTiffEncode; the error path then frees the output buffer without closing or clearing the TIFF handle, and later cleanup flushes through the write callback into freed memory, which can crash the process and may enable code execution. This issue is fixed in version 13.0.0.
Pillow is a Python imaging library. From 8.2.0 until 13.0.0, Pillow's tiled-TIFF decoder can read past the tile buffer while copying four-byte pixels. In _decodeTile, libtiff can return a tile allocation smaller than the required row stride multiplied by the tile length, and Pillow's stride-and-copy loop walks beyond that allocation. An attacker can supply a crafted tiled TIFF to an application that loads it. This can expose adjacent heap data in pixels or crash the process. This issue is fixed in version 13.0.0.
Pillow is a Python imaging library. Prior to 13.0.0, Pillow's P;2L and P;4L raw decoders can read past the end of an input buffer. ImagingRawDecode calculates a packed-pixel row size, while the unpackers read separately padded bit planes. A caller can supply a buffer and image dimensions whose accepted size is smaller than the unpacker's actual footprint. This can disclose up to three adjacent heap bytes in output pixels or crash the process. This issue is fixed in version 13.0.0.
Pillow is a Python imaging library. From 2.0.0 until 13.0.0, Pillow has a heap buffer overflow and stale-heap disclosure in _tiffWriteProc when saving compressed TIFF data through libtiff to a target without a usable fileno(), such as BytesIO. The callback sizes its allocation from the prior buffer size instead of the current write position, allowing an attacker-controlled out-of-line TIFF tag or retained ICC profile to be copied beyond the allocation after libtiff seeks forward. This overflow can crash the process and may allow code execution. Independently, unwritten alignment gaps in compressed TIFF output larger than 64 KiB can disclose stale heap bytes without a crafted tag. File paths and targets with real file descriptors are unaffected. This issue is fixed in version 13.0.0.
Pillow is a Python imaging library. Prior to 13.0.0, Pillow's rank-based filters can dereference invalid image storage for a zero-size image. ImagingExpand() allocates a non-empty output and then faults while replicating edges by reading the first pixel from empty input storage; ImagingRankFilter is never reached. An attacker can supply or cause creation of a zero-size image to an application that applies a rank, median, minimum, or maximum filter. This causes a native process crash. This issue is fixed in version 13.0.0.
Pillow is a Python imaging library. From 5.1.0 until 13.0.0, Pillow's PdfParser.get_value() can recurse without a limit while reading nested PDF objects. The parser follows attacker-controlled arrays and dictionaries when an existing PDF is opened for append. An attacker can provide a deeply nested PDF to an application that saves an image with PDF append enabled. This can raise an uncaught RecursionError or terminate the worker, causing denial of service. This issue is fixed in version 13.0.0.
Pillow is a Python imaging library. From 7.2.0 until 13.0.0, Pillow can enter an infinite loop while determining the frame count of a MIC image. MicImageFile inherits TIFF frame traversal, but MicImageFile.tell reports the OLE stream index and its seek() resets TIFF frame state, so an embedded TIFF with at least three IFDs never advances. When an application with olefile installed reads n_frames or saves all frames from a crafted MIC file, the non-advancing traversal can hold a worker at full CPU indefinitely. This issue is fixed in version 13.0.0.
The freeipmi packages contain an Intelligent Platform Management Interface (IPMI) remote console and system management software based on the IPMI specification.
Security Fix(es):
* FreeIPMI: FreeIPMI: Arbitrary code execution via malformed Fujitsu System Event Log responses (CVE-2026-85504)
* FreeIPMI: FreeIPMI: Arbitrary code execution via stack-based buffer overflow (CVE-2026-85507)
* FreeIPMI: FreeIPMI: Arbitrary code execution via stack-based buffer overflow in ipmi-oem (CVE-2026-85506)
* FreeIPMI: FreeIPMI: Arbitrary code execution via stack-based buffer overflow (CVE-2026-85509)
* FreeIPMI: FreeIPMI: Denial of Service via stack-based buffer over-read in ipmi-oem (CVE-2026-85505)
* freeipmi: FreeIPMI: stack-based buffer overflow in ipmi-oem Dell system info handler (CVE-2026-85508)
Bug Fix(es) and Enhancement(s):
* Update freeipmi to 1.6.20 to fix two potential buffer overflows [rhel-9.8.z] (JIRA:RHEL-279523)
For more details about the security issue(s), including the impact, a CVSS score, acknowledgments, and other related information, refer to the CVE page(s) listed in the References section.
MLflow 3.17.1 and earlier contains a path traversal vulnerability in the FileStore tracking backend that allows unauthenticated attackers to control filesystem paths by submitting a crafted trace ID containing traversal sequences to the POST /api/3.0/mlflow/traces endpoint. Attackers can exploit FileStore.start_trace() to write trace_info.yaml, request_metadata, and tag files with attacker-controlled metadata outside the intended experiment trace directory, subject to the server process's filesystem permissions. FileStore is the default tracking backend before MLflow 3.7.0 and opt-in thereafter (the default changed to a SQLite database), so deployments in the earlier part of the affected range are vulnerable without special configuration while later ones are affected only when FileStore is explicitly selected.
A weakness has been identified in NocoDB up to 2026.07.0. The affected element is the function validatePassword of the file /auth/user/signup. Executing a manipulation can lead to improper restriction of excessive authentication attempts. It is possible to launch the attack remotely. This attack is characterized by high complexity. The exploitability is described as difficult. The vendor documents this as intentional. The vendor deleted the GitHub issue for this vulnerability without any explanation. Afterwards the vendor was contacted early about this disclosure via email but did not respond in any way.
CVE-2026-116122 is an information disclosure vulnerability in NocoDB version 2026.07.0 affecting the AuthService.validateUser function in the Sign-in Endpoint. The flaw allows remote attackers to manipulate the salt argument, potentially exposing sensitive information. The attack complexity is high, and exploitability is difficult. The vendor has not responded to disclosure attempts and deleted the related GitHub issue. No patch or official fix has been announced.
@apexacc/cli is distributed by "Apex Accelerator" (Apex Arena), a fake Web3 accelerator that lures crypto founders with free AI credits and audits. Elastic Security Labs (REF8801, published 2026-10-12) found it is a repackaged Oh My Pi coding agent (1,351 of 1,363 files identical) with malicious additions that hijack the developer's AI coding agent: it overwrites OPENROUTER_BASE_URL so model traffic (prompts, files read, tool results) goes to attacker infrastructure at arena.apexaccs.org; persists an attacker-controlled remote MCP server in ~/.apex/agent/mcp.json; and defaults tool approval to "yolo". A bootstrap retrieves commands from vrf.apexaccs.org: on macOS it drops CloudSyncD (fake password prompt, credential theft, C2 ccmihr.com); on Windows a staged loader chain ends in an infostealer that exfiltrates browser credentials, wallets, messengers, VPN clients and Claude/Cursor/Windsurf configs over Telegram. Elastic analyzed version 1.6.1 and traced the bootstrap to the earliest preserved version (script host arenapex.cloud). The package was still live on npm (latest 1.6.2) with no OSV record as of 2026-10-16. Independently confirmed by static inspection of the live 1.6.2 npm tarball (sha256 7ca5e3908dd8721d0cf6af2488f0825c0d91a5e0b7c8529ef830141a4a3ba31f) on 2026-10-16, without execution: the bin launcher apex.cjs (sha256 4029192a44ee738deba9aa53b57550075cb8fd170a63f0636dea5452256f2175) runs when the user invokes the CLI (no install script). On Windows it sets the current user's PowerShell execution policy to RemoteSigned (no elevation needed), and on Windows 11 it launches a hidden, elevated (UAC RunAs) PowerShell that attempts to add a Defender exclusion for ~/.apex and to set HKLM\SYSTEM\CurrentControlSet\Control\CI\Policy VerifiedAndReputablePolicyState=0 (disabling Smart App Control). From 1.5.122 (2026-09-17) through 1.6.2 those two commands are built in a JavaScript template literal with single backslashes, so the shipped paths are mangled (the exclusion lands on a sibling path '<home>.apex' and the registry key is invalid) and both likely fail; the UAC prompt still appears. Versions 1.5.120 (registry change) and 1.5.121 (adds the Defender exclusion), both 2026-09-17, carried correctly escaped versions of these commands without an elevation prompt, effective only when already running as administrator. The execution-policy change first appears in 1.6.2. It then downloads a ~114MB native binary from GitHub releases of apexaccs/copilot into ~/.apex/bin, and on macOS an additional apex-verify-darwin helper. Both GitHub repositories referenced by the package versions (apexaccs/copilot, Apex-Accelerator/apexcli) return 404 as of 2026-10-16. All versions affected. Indicators from Elastic's report.
---
_-= Per source details. Do not edit below this line.=-_
## Source: amazon-inspector (c174007782646e1fdfd9806ecd699e427fd36661dd7137fa462672f41aee6e79)
On CLI startup (invoked via npx per the package README), apex.cjs unconditionally detects Windows 11 and launches a hidden, UAC-elevated PowerShell via Start-Process -Verb RunAs -WindowStyle Hidden with a UTF-16LE base64 -EncodedCommand payload. The elevated payload (a) calls Add-MpPreference -ExclusionPath on %USERPROFILE%\.apex to whitelist the package directory in Windows Defender, and (b) writes HKLM\SYSTEM\CurrentControlSet\Control\CI\Policy VerifiedAndReputablePolicyState=0 via reg add, disabling Smart App Control / Code Integrity enforcement machine-wide. The elevation path uses underscore-prefixed identifiers (_ps, _enc, _sp, _osv, _isWin11, _home), a hidden window, base64 command encoding, and a broad try/catch that silently swallows errors. After weakening the host, the CLI downloads a ~105MB apex-<platform>.exe (and an optional native.node addon) from the publisher's GitHub Releases with no hash or signature verification, chmods 0755, and executes it via spawnSync with inherited stdio and the full process.env. The security-policy downgrade is not consistent with any documented function of an AI CLI and converts an otherwise routine publisher-pinned binary fetch into execution of unverified code under a machine-wide-weakened Windows security posture.
CVE-2026-116121 is a race condition vulnerability in NocoDB affecting version 2026.08.0. The flaw exists in the processing of the file users.service.ts within the Registration component. Exploitation requires a high level of complexity and is considered difficult. The vulnerability can be exploited remotely without user interaction or privileges. The vendor has not provided a response or patch and deleted the related GitHub issue without explanation.