VulnerabilityPublished 22 Oct 2026Updated 29 Oct 2026

CVE-2026-112898: Pillow · Image.quantize() with empty palette leaks uninitialized stack memory

Affects aiderSource GCVE Database

Compare

Authoritative record:CVE-2026-112898· Medium 6.9

Not known to be exploited · CVSS 6.9 Medium.

Data as of 30 Oct 2026, 07:30 UTC

CVSS 4.0 base score

6.9Medium

How severe the flaw is on the 0–10 CVSS scale.

Thin evidence

One scorer or none

CVSS scores by source
CNA (GitHub_M)	CVSS 4.0	6.9

CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:L/VI:N/VA:N/SC:N/SI:N/SA:N

Attack Vector: Network
Attack Complexity: Low
Attack Requirements: None
Privileges Required: None
User Interaction: None
Vulnerable system confidentiality: Low
Vulnerable system integrity: None
Vulnerable system availability: None
Subsequent system confidentiality: None
Subsequent system integrity: None
Subsequent system availability: None

Bold values are the worst case: easiest to exploit or largest impact.

Attack Vector: Network: The vulnerable system is bound to the network stack and the set of possible attackers extends beyond the other options listed below, up to and including the entire Internet. Such a vulnerability is often termed “remotely exploitable” and can be thought of as an attack being exploitable at the protocol level one or more network hops away (e.g., across one or more routers).
Attack Complexity: Low: The attacker must take no measurable action to exploit the vulnerability. The attack requires no target-specific circumvention to exploit the vulnerability. An attacker can expect repeatable success against the vulnerable system.
Attack Requirements: None: The successful attack does not depend on the deployment and execution conditions of the vulnerable system. The attacker can expect to be able to reach the vulnerability and execute the exploit under all or most instances of the vulnerability.
Privileges Required: None: The attacker is unauthorized prior to attack, and therefore does not require any access to settings or files of the vulnerable system to carry out an attack.
User Interaction: None: The vulnerable system can be exploited without interaction from any human user, other than the attacker.
Vulnerable system confidentiality: Low: There is some loss of confidentiality. Access to some restricted information is obtained, but the attacker does not have control over what information is obtained, or the amount or kind of loss is limited. The information disclosure does not cause a direct, serious loss to the Vulnerable System.
Vulnerable system integrity: None: There is no loss of integrity within the Vulnerable System.
Vulnerable system availability: None: There is no impact to availability within the Vulnerable System.
Subsequent system confidentiality: None: There is no loss of confidentiality within the Subsequent System or all confidentiality impact is constrained to the Vulnerable System.
Subsequent system integrity: None: There is no loss of integrity within the Subsequent System or all integrity impact is constrained to the Vulnerable System.
Subsequent system availability: None: There is no impact to availability within the Subsequent System or all availability impact is constrained to the Vulnerable System.

Definitions: FIRST CVSS v4.0 specification.

Exploitation evidence

Disclosed: evidence found.
CVE published 19 Oct 2026
Detectable: no evidence on record.
No public nuclei template indexed
Weaponized: no evidence on record.
No public exploit code indexed in Metasploit or Exploit-DB
Exploited in the wildExploited: no evidence on record.
Not in CISA KEV and no source reports exploitation
Observed by OffSeq MirageSeen by Mirage: no evidence on record.
No attack traffic recorded by OffSeq Mirage sensors

A missing step is not evidence of safety. Sources: cve.org (opens in a new tab), the Metasploit, Exploit-DB and nuclei-templates index as of 29 Oct 2026.

Disclosed: The CVE record is public on cve.org. Its publication is t = 0.
Detectable: A public nuclei template can find exposed instances, for defenders and attackers alike.
Weaponized: A Metasploit exploit module or Exploit-DB entry exists (Metasploit scanner modules do not count), or CISA SSVC or the CVSS exploit maturity reports a public proof of concept.
Exploited in the wild: CISA KEV lists it, CISA SSVC rates exploitation active, or a source reports attacks.
Observed by OffSeq Mirage: OffSeq Mirage honeypot sensors recorded attack traffic for this CVE, or found exposed hosts without attack traffic.

Moments are spaced evenly, not to scale; a dashed line marks a gap of six months or more. Deltas count from t = 0. Sources that give only a date (CISA KEV, Exploit-DB, or a time of exactly 00:00Z) read “same day” rather than hours. Metasploit modules carry the flaw’s disclosure date, not their release date, and nuclei templates carry no date, so both appear above but not on the timeline.

What to do

Affected software

Product	Affected	Fixed in
Homebrewaider`pkg:brew/aider`Other threats	Affected >=0.86.1_1	Fixed in–

Check your own hosts for these packages with threat-finder (opens in a new tab), our free open-source scanner.

Description

``` ============================================================================ Image.quantize() with empty palette leaks uninitialized stack memory Location: src/libImaging/Palette.c:216 in ImagingPaletteCacheUpdate() Commit: a707c6b21c313880d0771e54965d142dd42358b6 ============================================================================ ``` When an application lets a user supply the palette image for Pillow's `Image.quantize(palette=...)`, the user can get back 512 bytes of the server process's stack memory in the pixels of the returned image. An 80-byte PNG with an empty palette is enough, and the leaked bytes survive saving to PNG or GIF. In the runs examined, those bytes included memory addresses inside the stack, heap, libc, libpython and Pillow's own C module, along with the stack-protector canary. The bug is in `ImagingPaletteCacheUpdate()` in `src/libImaging/Palette.c`. It fills a local `UINT8 c[512]` only inside loops bounded by `palette->size`. When the palette has zero entries, the array is never written, but it is still copied into the colour cache. `topalette()` in `src/libImaging/Convert.c` then emits the cached values as output pixel indices. `topalette()` rejects a `NULL` palette but not one with `size == 0`. Three inputs produce such a zero-entry `"P"` image: a PNG with a zero-length `PLTE` chunk opened with `Image.open()`, a colour-mapped TGA with colour-map length 0, or a bare `Image.new("P", ...)`. The attacker must control the *palette* image; the source image can be anything, and `RGB` sources are affected with either dither mode. There is no out-of-bounds access and no memory corruption. The consequence is a bounded stack disclosure that can reveal address layout and the canary of a long-lived worker, with no crash and no log entry. The attached reproducer quantizes a 512-colour strip against the empty-palette PNG and finds 8 of the 64 output words pointing into the process's own `[stack]` mapping, while a one-entry palette control yields all zeros. The behaviour was introduced in commit `54cb09d8b48c` (first released in 9.1.0). The `Image.new("P", ...)` route exists since 10.1.0. ## Details `ImagingPaletteCacheUpdate()` (`src/libImaging/Palette.c:154`) computes, for one 8×8×8 box of the colour cache, the nearest palette index for each of the 512 slots into a local `UINT8 c[512]` and then copies `c[]` into the cache. `c[]` is only ever written inside `for (i = 0; i < palette->size; i++)`. When the palette has zero entries nothing writes it, the untouched stack bytes are copied into the cache, and `topalette()` (`src/libImaging/Convert.c:1145`) emits them as the output pixels of `Image.quantize(palette=...)`. The root cause is that `topalette()` only rejects a `NULL` palette, never one with `palette->size == 0`, while `ImagingPaletteCacheUpdate()` silently assumes `size >= 1`. A zero-entry `"P"` image is what `Image.open()` produces for a PNG whose `PLTE` chunk has length 0 (or a colour-mapped TGA with colour-map length 0); `Image.new("P", size)` without a subsequent `putpalette()` yields one as well. Both dither modes (`Dither.NONE` and the default Floyd-Steinberg) are affected for `RGB` sources (an `L` source takes the grayscale copy path and never consults the cache). Showing it with the installed build, no special flags. The 80-byte PNG below is 4×4, bit depth 8, colour type 3, with a zero-length `PLTE` and an all-zero `IDAT`: ``` 89504e470d0a1a0a 0000000d49484452 0000000400000004 0803000000 9e2f6e4c 00000000 504c5445 4ba88955 0000000b 49444154 789c6360c0040000140001 7dfe1eee 00000000 49454e44 ae426082 ``` ```python pal = Image.open(io.BytesIO(png)); pal.load() # pal.mode == "P", pal.im.getpalette("RGB") == b"" src = Image.new("RGB", (512, 1)) src.putdata([(r*4, g*4, b*4) for r in range(8) for g in range(8) for b in range(8)]) out = src.quantize(palette=pal, dither=Image.Dither.NONE) out.tobytes() # 512 bytes, different in every process ``` The 512 source colours are chosen so that each pixel lands in a distinct slot of the cache box for `r,g,b ∈ [0,32)`, in slot order, so `out.tobytes()` is exactly `c[0..511]`. Read as 64 little-endian 8-byte words and resolved against the same process's `/proc/self/maps`, a typical run gives 43 words that are addresses inside the process's own mappings — 8 in `[stack]`, several in `[heap]`, `libc.so.6`, `libpython3.12.so` and `_imaging.cpython-312-x86_64-linux-gnu.so` — and 3 words equal to the thread's stack-protector canary (read from `fs:0x28` for comparison). The same PNG with a 3-byte `PLTE` (one entry) gives 512 zero bytes in every run. `out.save(buf, "PNG")` succeeds, so the bytes leave the process in an ordinary image. Under valgrind (run against a `strip --strip-debug` copy of `_imaging.so` or a gcc build, since valgrind 3.19 cannot read clang's DWARF5) the report carries the origin `Uninitialised value was created by a stack allocation at ImagingPaletteCacheUpdate`. With `Dither.NONE` `topalette` only copies

References

Community reviews

Loading reviews…