--- id: "CVE-2026-112901" kind: "vulnerability" aliases: ["CVE-2026-112901", "GHSA-xgh8-jp6q-p5x8"] title: "Pillow: ReDoS in XbmImagePlugin: 512-byte file stalls Image.open() for minutes" severity: "medium" cvss: 6.9 epss: 0.0029100000392645597 exploited: false status: "open" category: "package" products: [{"name": "Pillow", "fixed": ["13.0.0"], "affected": "<= 12.3.0", "ecosystem": "pypi"}] ecosystems: ["pypi"] tags: ["cwe-1333", "image-upload", "package", "pillow", "python", "thumbnailer"] not_affected_if: ["formats-allowlist-excludes-xbm", "no-untrusted-image-input", "pillow-13.0.0-or-later"] published_at: "2026-10-19T22:17:10+00:00" first_seen_at: "2026-10-23T19:11:32+00:00" updated_at: "2026-10-23T21:57:19+00:00" sources: [{"source": "ghsa", "url": "https://github.com/advisories/GHSA-xgh8-jp6q-p5x8", "first_seen_at": "2026-10-23T19:11:32+00:00"}, {"source": "nvd", "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-112901", "first_seen_at": "2026-10-23T21:57:19+00:00"}] --- # Pillow: ReDoS in XbmImagePlugin: 512-byte file stalls Image.open() for minutes Pillow's XBM image plugin has a regular expression that takes minutes of CPU to reject a crafted 512-byte file, and Image.open() runs it on any byte stream that starts with #define. An attacker who can submit bytes to an upload endpoint or thumbnailer can tie up a worker process for minutes per request. _Summary, check and fix written by a model from the sources; labelled as such._ ## Affected - pypi:Pillow affected <= 12.3.0 fixed 13.0.0 ## How to check pip show pillow (affected if version is 12.3.0 or lower) If the app calls Image.open() without a formats= allowlist that excludes "XBM", it is exposed grep -rn "Image.open" to find call sites that accept untrusted input ## Fix Upgrade Pillow to 13.0.0 or later Until upgraded, pass formats=[...] to Image.open() that excludes "XBM" ## Timeline - 2026-10-23T19:11:32+00:00 ghsa https://github.com/advisories/GHSA-xgh8-jp6q-p5x8 - 2026-10-23T21:57:19+00:00 nvd https://nvd.nist.gov/vuln/detail/CVE-2026-112901 ## From the sources ``` ============================================================================ ReDoS in XbmImagePlugin: 512-byte file stalls Image.open() for minutes Location: src/PIL/XbmImagePlugin.py:58 in XbmImageFile._open() Commit: a707c6b21c313880d0771e54965d142dd42358b6 ============================================================================ ``` A crafted 512-byte file makes Pillow's `Image.open()` spend about six minutes of processor time on one core before it gives up with `UnidentifiedImageError`. Anyone who can hand bytes to an application that opens images with Pillow's default settings can trigger this, for example through an upload form, a thumbnailer, or a service that checks whether a blob is an image. A few such files sent at once can tie up a typical pool of worker processes. The cause is the `xbm_head` regular expression in `src/PIL/XbmImagePlugin.py:32-40`, which `XbmImageFile._open()` (line 58) runs over the first 512 bytes of any stream whose prefix starts with `#define`. Several adjacent parts of the pattern can match the same characters. `[ \t]+` overlaps with `.*` in the width and height lines and with `[^_]*` in the optional hotspot lines, and `[\r\n]+` closing the hotspot group overlaps with the following `[\000-\377]*`. The header sets a run of k spaces after each `#define` and a run of k line terminators after `_y_hot`, then omits the `[]` after `_bits`. That forces Python's backtracking matcher to try every way of splitting those runs and rescan the remaining bytes each time. The cost grows roughly as k to the fifth power: k = 60 measured 375 s, while a well-formed XBM of the same size opens in about 40 ms. Because this happens before any format is identified, none of Pillow's size, pixel-count or decompression-bomb limits apply, and upload size limits do not help because the file is tiny. `ImageFile.Parser.feed()` repeats the cost on every chunk until the stream is identified. The only existing mitigation is passing a `formats=` allowlist that excludes `"XBM"`. The attached reproducer opens the same header with and without the trailing `[]`. The valid version opens as XBM in 52 ms. The broken version takes 0.1 s, 0.75 s, 3.0 s and 8.6 s at k = 10, 15, 20 and 25, and the slow k = 60 case is not run. This is the same class of bug as CVE-2021-25292 in `PdfParser`, but in a different plugin and reachable from plain `Image.open()`. ## Details `XbmImageFile._open()` identifies an XBM file by running the module-level regex `xbm_head` over the first 512 bytes of the stream. The pattern has several adjacent quantifier pairs that can consume the same characters, followed by a `[\000-\377]*_bits\[]` tail that rescans the rest of the buffer for every way of splitting them. A 512-byte input built to exercise all of them and then fail the final literal keeps `Image.open()` in `xbm_head.match()` for minutes before it raises `UnidentifiedImageError`. This 512-byte input (k = 60) takes 375 s of CPU in `Image.open(io.BytesIO(data))` with default settings, measured on the stock build at this commit: ```python k = 60 data = b"#define" + b" "*k + b"_width 1\r" data += b"#define" + b" "*k + b"_height 1\r" data += b"#define" + b" "*k + b"_x_hot 1\r" data += b"#define" + b" "*k + b"_y_hot 1" + b"\r"*k data += b"_bits" # no "[]": the final literal never matches data += b"A" * (512 - len(data)) # pad to the 512 bytes _open() reads ``` Calling `XbmImagePlugin.xbm_head.match(data)` directly gives the same picture: k = 10 → 0.11 s, k = 15 → 0.76 s, k = 20 → 3.0 s, k = 25 → 8.5 s, k = 30 → 19.9 s (roughly ×27 per doubling of k, i.e. about k^4.7). A well-formed 512-byte XBM matches in 10 µs and opens and loads through `Image.open()` in ~40 ms including plugin import; the same k = 60 header with `_bits[]` present matches in 5 µs. The regex, `src/PIL/XbmImagePlugin.py:32`: ```python xbm_head = re.compile( rb"\s*#define[ \t]+.*_width[ \t]+(?P[0-9]+)[\r\n]+" b"#define[ \t]+.*_height[ \t]+(?P[0-9]+)[\r\n]+" b"(?P" b"#define[ \t]+[^_]*_x_hot[ \t]+(?P[0-9]+)[\r\n]+" b"#define[ \t]+[^_]*_y_hot[ \t]+(?P[0-9]+)[\r\n]+" b")?" rb"[\000-\377]*_bits\[]" ) ``` and the only call site, `src/PIL/XbmImagePlugin.py:58`: ```python def _open(self) -> None: … m = xbm_head.match(self.fp.read(512)) if not m: msg = "not a XBM file" raise SyntaxError(msg) ``` For a well-formed header (`#define im_width 8\n…static char im_bits[] = {`) each `[ \t]+` eats the single space, `.*`/`[^_]*` eat the identifier prefix, and `[\000-\377]*_bits\[]` finds the literal on its first backward scan, so the match is linear. The crafted header diverges at every one of those points: - `[ \t]+.*_width` and `[ \t]+.*_height`: the run of k spaces after `#define` can be split between `[ \t]+` and `.*` in k ways, all leading to the same continuation. Factor k each. (The choice of `\r` terminators is incidental: the same header with `\n` terminators is equally slow.) - `[ \t]+[^_]*_x_hot` and `[ \t]+[^_]*_y_hot` inside the optional `hotspot` group: `[^_]*` also matches spaces, so the same k-way split exists twice more. Factor k each. These two lines are essential: without them the same input is rejected in about 10 ms. - `[\r\n]+` closing the `hotspot` group, followed by `)?` and `[\000-\377]*`: the run of k `\r` bytes after `_y_hot 1` can be divided between `[\r\n]+` and `[\000-\377]*` in k ways. Factor k. - `[\000-\377]*_bits\[]`: for every combination above, the engine swallows the remaining ~150 bytes of padding and walks back looking for `_bits[]`. The input contains `_bits` but never `[]`, so this always fails and every combination is visited. CPython's `re` is a backtracking engine without memoisation, so the total work is on the order of k^5 × (512 − header length) steps — about 10^11 for k = 60, consistent with the measured 375 s. The 512-byte read caps k (5k + ~60 ≤ 512), so the cost per `open()` is bounded, but the bound is minutes of CPU. The path is the default one. `Image.open()` (`src/PIL/Image.py:3719`) reads a 16-byte prefix and offers it to every registered plugin's accept hook; `XbmImagePlugin._accept()` (`src/PIL/XbmImagePlugin.py:43`) returns true for anything whose left-stripped prefix starts with `#define`, and `_open()` then runs the regex. No size, pixel-count or decompression-bomb check can apply before this point because nothing has been parsed yet, and since the match ultimately fails the file is never identified, so the only outcome for the caller is `UnidentifiedImageError` after the delay. File name and extension are irrelevant because dispatch is by content. A `faulthandler` stack dump taken 20 s into the call shows the interpreter at `XbmImagePlugin.py:58 in _open` ← `ImageFile.py:154 in __init__` ← `Image.py:3699 in _open_core` ← `Image.py:3719 in open`. `ImageFile.Parser.feed()` (`src/PIL/ImageFile.py:584`) amplifies this further: until the stream is identified it calls `Image.open()` on all accumulated data at every `feed()`, so an application feeding the stream in chunks pays the regex cost once per chunk (measured: k = 15 header fed in 8 chunks → 4.6 s versus 0.76 s for a single `open()`). ## Impact Anyone who can get 512 bytes in front of `Image.open()` — an upload endpoint, thumbnailer, or any service that uses Pillow to sniff whether a blob is an image — pins one CPU core of the handling process for roughly six minutes per request with the default plugin set and default limits. Nothing large is allocated and no `DecompressionBombWarning`/`Error` fires; application-side upload size limits do not help because the file is tiny, and `Image.MAX_IMAGE_PIXELS` is never consulted because no image is ever identified. A handful of concurrent requests is enough to exhaust a typical worker pool. The per-call cost is bounded by the 512-byte header read, so this is a CPU-exhaustion denial of service of the handling process --- _Free to read at a human pace. An API with keys and an MCP server are in preparation; early access: hello@infrastack.ai_