yara-x: Unvalidated deserialization in safe `Rules::deserialize` allows memory corruption and UB

CISA KEV is observed exploitation evidence. FIRST EPSS is a dated prediction of exploitation in the next 30 days. Missing scores are unknown. This collector requests predictions only for explicit CVE aliases in the public NuGet archive; Microsoft release views retain their existing prediction scope.

No collected risk observation for these identifiers. This does not establish absence of exploitation or a zero probability.

> [!NOTE] > This finding was identified during an agentic unsafe Rust code review performed by Gemini AI, followed by human review and verification. ## The Issue The crate exports a public safe API [`Rules::deserialize`](https://github.com/VirusTotal/yara-x/blob/5bd1f35db783679c90a3ea1a66bd15fe4e55bef1/lib/src/compiler/rules.rs#L187-L254) accepting any generic byte sequence `B: AsRef<[u8]>`. It restores compiled rule structures directly from raw bytes using `bincode::serde::decode_from_slice`. This decoded `Rules` struct contains internal lookup tables, including `sub_patterns: Vec<(PatternId, SubPattern)>`, `atoms: Vec<SubPatternAtom>`, and `lit_pool: BStringPool`. Subsequent safe operations assume these internal tables satisfy strict structural invariants: - `Rules::get_sub_pattern` executes `unsafe { self.sub_patterns.get_unchecked(sub_pattern_id.0 as usize) }`. If untrusted serialized bytes contain an atom referencing an out-of-bounds `SubPatternId`, calling `get_sub_pattern` during scanning triggers an out-of-bounds memory read (Undefined Behavior). https://github.com/VirusTotal/yara-x/blob/5bd1f35db783679c90a3ea1a66bd15fe4e55bef1/lib/src/compiler/rules.rs#L355-L360 - `Metadata::next()` extracts string metadata via `unsafe { s.to_str_unchecked() }`. If serialized bytes corrupt `lit_pool` indices or structural data, `to_str_unchecked` constructs a `&str` pointing to invalid UTF-8 bytes (Undefined Behavior). https://github.com/VirusTotal/yara-x/blob/5bd1f35db783679c90a3ea1a66bd15fe4e55bef1/lib/src/models.rs#L204-L210 Because passing malformed or untrusted data to `Rules::deserialize` induces Undefined Behavior in subsequent safe calls (`Scanner::new`, `Scanner::scan`) without any `unsafe` blocks in caller code, this API is unsound. <details><summary>Minimal Reproduction (Miri / Native Crash)</summary> Zip file with crashing_payload: [crashing_payload.zip](https://github.com/user-attachments/files/29173221/crashing_payload.zip) We have a payload crashing_payload.bin where only a single byte in the structural metadata tail is mutated (changing a `SubPatternId` from `1` to `248` while keeping the WebAssembly bytecode completely untouched and valid). Below is the self-contained verification script which compiles and runs against the official **unmodified** `yara-x v1.17.0` crate: ```rust use yara_x::{Rules, Scanner}; fn main() { // Embed the crashing payload generated by the fuzzer at compile time. let serialized = include_bytes!("crashing_payload.bin"); println!("Loaded embedded crashing payload, length: {}", serialized.len()); // Deserialize. On unmodified library, this succeeds because the WASM and headers // are pristine and structural corruption isn't validated. if let Ok(deserialized) = Rules::deserialize(serialized) { println!("Deserialization succeeded! Running scanner..."); let mut scanner = Scanner::new(&deserialized); // Run the standard scan, which will execute the WASM and trigger the out-of-bounds read! let _ = scanner.scan(b"lorem ipsum dolor sit amet"); println!("Scanner finished."); } else { println!("Deserialization failed!"); } } ``` ### 1. Miri Trace NOTE: This needs to be run with 1.17.0. I haven't tested this against other versions. Unfortunately I was able to get a miri trace, but I'm not able to reproduce it right now because of lockfile changes. If you're trying this out be sure to use `MIRIFLAGS="-Zmiri-disable-stacked-borrows"` ### 2. Segfault / panics When run natively (without Miri or any sanitizers) on a standard Linux platform, the process immediately segfaults: ```bash $ cargo run --bin verify Loaded embedded crashing payload, length: 11777 Deserialization succeeded! Running scanner... Segmentation fault (core dumped) ``` And with a newer compiler (which appears to have debug assertions in `get_unchecked)` ```bash Deserialization succeeded! Running scanner... thread 'main' (997442) panicked at lib/src/compiler/rules.rs:404:36: unsafe precondition(s) violated: slice::get_unchecked requires that the index is within the slice This indicates a bug in the program. This Undefined Behavior check is optional, and cannot be relied on for safety. note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace ``` </details> <details><summary>Suggested Fix</summary> To uphold Rust soundness guarantees, either mark `Rules::deserialize` as `pub unsafe fn deserialize` with a formal `/// # Safety` contract documenting that callers are responsible for verifying the authenticity and structural integrity of the input bytes (e.g. via cryptographic signatures), or replace all internal `get_unchecked` and `to_str_unchecked` calls on deserialized data structures with safe bounds checks (`.get()`) and UTF-8 validation (`std::str::from_utf8`). </details> ---

Source published

24 Sep 2026 19:09:28 UTC

Source modified

24 Sep 2026 19:15:05 UTC

Observed here

24 Oct 2026 13:19:54 UTC

Source terms

GitHub Advisory Database · CC-BY-4.0

Original source record · Cite this exact observation · Complete original publisher JSON

This view projects the publisher's source fields for reading; the original JSON retains additional fields and CVE enrichment containers. Attribution: osv · GitHub Advisory Database · CC-BY-4.0. Source text is retained; field grouping and change labels are provided by ChangeIntel.

CVSS_V4 · CVSS:4.0/AV:L/AC:L/AT:N/PR:L/UI:N/VC:N/VI:N/VA:L/SC:N/SI:N/SA:N

Publisher rating · MODERATE