Skip to content

vhdx-forensic

Read and audit Hyper-V VHDX disk images in pure Rust — a hardened Read + Seek container reader plus a 63-code structural integrity analyzer for DFIR.

This workspace ships two crates: vhdx-core — the VHDX/Hyper-V virtual-disk container reader (dynamic, fixed, differencing, automatic dirty-log recovery), exposing a Read + Seek view over the virtual sector stream (published as vhdx-core, imported as vhdx); and vhdx-forensic — the integrity analyzer that audits the raw bytes for tampering, corruption, and anti-forensic GUID/log wiping, emitting forensicnomicon::report::Finding plus optional in-memory CRC repair. Zero unsafe code, no C bindings, no external tools.

[dependencies]
vhdx-core = "0.2"       # reader — imported as `vhdx`
vhdx-forensic = "0.2"   # analyzer — graded structural findings
use vhdx_forensic::{anomalies_at_least, Severity, VhdxIntegrity};

let image = std::fs::read("disk.vhdx")?;
let anomalies = VhdxIntegrity::new(&image).analyse();
for a in anomalies_at_least(&anomalies, Severity::Error) {
    println!("[{:?}] {}", a.severity(), a.forensic_significance());
}
# Ok::<(), Box<dyn std::error::Error>>(())

Supported formats

Format Supported
VHDX Version 1 (Windows 8 / Server 2012+)
Dynamic disks (sparse, BAT-addressed)
Fixed disks (pre-allocated)
Differencing disks (single-level parent chain)
Log replay (dirty-log recovery)

Read-only. Differencing disks require the parent image to be supplied via VhdxReader::from_bytes_with_parent. Log replay is applied automatically on open when the active header carries a non-zero LogGuid.

Forensic analysis — vhdx-forensic

VhdxIntegrity::new(&bytes).analyse() walks the raw container across six phases — container/magic, CRC integrity, header semantics, region layout, metadata, and BAT/data-block analysis. There are 63 distinct anomaly codes; a representative sample:

Code Severity Meaning
VHDX-BAD-MAGIC Critical File does not start with the vhdxfile signature
VHDX-BOTH-HEADER-COPIES-INVALID Critical Both header copies fail CRC32C — header recovery impossible
VHDX-HEADER-CHECKSUM-MISMATCH Error A header copy's stored CRC32C does not match the computed value
VHDX-REGIONS-OVERLAP Error Two region-table entries claim overlapping container bytes
VHDX-LOG-ENTRY-CRC-MISMATCH Error A log entry's CRC fails — replay would corrupt the image
VHDX-BAT-ENTRY-BEYOND-CONTAINER Error A BAT entry points past the end of the file
VHDX-FILE-WRITE-GUID-ALL-ZEROS Warning FileWriteGuid wiped — consistent with anti-forensic tampering
VHDX-GHOST-DATA-IN-ABSENT-BLOCK Warning Non-zero payload in a block marked absent — residual/hidden data
VHDX-DIFFERENCING-DISK Warning Image is a child disk requiring a parent to read
VHDX-DIRTY-LOG Info Active log present — image was not cleanly closed

Findings are observations, never legal conclusions — MITRE mappings are surfaced as "consistent with," for the analyst to weigh. VhdxRepair::new(bytes).attempt_repair() rebuilds header and region-table CRC32C checksums from a valid peer copy in memory; it never alters payload data.

Trust but verify

Panic-free on hostile input (no unwrap/expect/panic! or unchecked indexing in production code, hard deny lints, bounds-checked length/offset/count fields, checked_mul/checked_add BAT addressing), fuzzed via a cargo-fuzz workspace, and validated against real artifacts from the log2timeline/dfvfs corpus and QEMU output with virtual disk sizes cross-checked against qemu-img info, with the decoded byte stream verified byte-identical to qemu-img convert -O raw. See Validation.

See the project README for the full usage guide, CLI, and related-crate map.


Privacy Policy · Terms of Service · © 2026 Security Ronin Ltd