snappy-java can be crashed by small crafted compressed data
A tiny specially crafted compressed input can make a Java program try to reserve up to 2 GB of memory, which can crash it. Only programs that decompress data from untrusted sources are at risk.
| Severity | HighCVSS 7.5 of 10 |
|---|---|
| Being exploited? | UnclearNo reports either way yet |
| Fix available? | Yes1.1.10.9 · since 27 Oct 2026 |
| How sure is this? | Official sourceBacked by an official source |
How the attack works simplified
- 1Crafted inputFew bytes declaring a huge length
- 2App decompressesCalls Snappy.uncompress or stream reader
- 3Large allocationBuffer sized from declared length
- 4Service failsOutOfMemoryError or exception
- ResultService crashes from memory exhaustion
What happened
In snappy-java, several decompression methods allocate an output buffer based on the uncompressed length declared in the first bytes of the input before validating it, enabling memory exhaustion and NegativeArraySizeException.
Are you affected?
You are affected if you use org.xerial.snappy:snappy-java version 1.1.10.8 or earlier and decompress data from untrusted sources.
| Package / product | Affected versions | Fixed in |
|---|---|---|
| org.xerial.snappy:snappy-java · maven | <= 1.1.10.8 | 1.1.10.9 |
What to do
- Update org.xerial.snappy:snappy-java to 1.1.10.9 or later
- If you cannot update yet, call Snappy.uncompressedLength(input) on untrusted data and reject inputs above an application-specific limit before decompressing
Timeline
- 27 Oct 2026, 15:33 UTCPublicly disclosed
- 27 Oct 2026, 15:33 UTCReported by GitHub Advisory · GitHub Advisory
- 27 Oct 2026, 15:33 UTCPicked up by Zero Radar
Sources 1 report, oldest first
Official vendor, government CERT or advisory database
- snappy-java: Unbounded memory allocation from attacker-declared uncompressed lengthaGitHub Advisory · 27 Oct 2026, 15:33 UTC
Technical details
- ID
- CVE-2026-108106 · GHSA-q233-g92q-qxp6
- CVSS
- 7.5 / 10
- EPSS
- 0.4% chance of exploitation within 30 days
- Disclosed
- 27 Oct 2026, 15:33 UTC
- On Zero Radar
- 27 Oct 2026, 15:33 UTC (0 min after disclosure)
- Last change
- 27 Oct 2026, 15:33 UTC
The plain-language parts and the attack diagram were written by an AI model (anthropic/claude-haiku-5.5) from the sources above. Facts like versions, exploitation and dates come from the sources themselves. How Zero Radar works