regipy

⚠️ Breaking Changes in v6.0.0

Version 6.0.0 includes significant modernization changes:

Python 3.9+ required - Dropped support for Python 3.6, 3.7, and 3.8

attrs library removed - Data classes now use Python's built-in dataclasses module

If your code imports internal classes (Cell, VKRecord, Value, Subkey) and uses attrs functions like attr.asdict(), switch to dataclasses.asdict()

See the CHANGELOG for full details.

Regipy is a python library for parsing offline registry hives (Hive files with REGF header). regipy has a lot of capabilities:

Use as a library:
- Recurse over the registry hive, from root or a given path and get all subkeys and values
- Read specific subkeys and values
- Apply transaction logs on a registry hive
Command Line Tools
- Dump an entire registry hive to json
- Apply transaction logs on a registry hive
- Compare registry hives
- Execute plugins from a robust plugin system (i.e: amcache, shimcache, extract computer name...)

Requires Python 3.9 or higher.

Installation

Regipy latest version can be installed from pypi:

pip install regipy[full]

NOTE: regipy[full] installs dependencies that require compilation tools and might take some time. It is possible to install a version with relaxed dependencies, by omitting the [full].

Also, it is possible to install from source by cloning the repository and executing:

pip install --editable .[full]

Rust-accelerated backend (alpha)

An optional Rust implementation of the core REGF parser is available as an opt-in backend, published separately to PyPI as regipy-rs:

pip install regipy[rust]

from regipy.registry_rs import RegistryHive  # instead of regipy.registry

reg = RegistryHive("/tmp/NTUSER.dat")
# Same API: get_key, iter_values, recurse_subkeys, plugins — everything
# works unchanged, including all regipy plugins.

It is a drop-in replacement validated 1:1 against the pure-Python parser over the entire test-hive corpus — every key path, timestamp, value and plugin output, notarized by matching SHA-256 traversal digests (see regipy_tests/comparison_test.py and the Forensic parity evidence section of regipy-rs/BENCHMARKS.md).

Full traversal with values (recurse_subkeys), best of 3 runs:

Hive	Keys	Python	Rust	Speedup
NTUSER.DAT	1,812	173 ms	5 ms	38x
UsrClass.dat	6,205	948 ms	17 ms	55x
amcache.hve	2,105	837 ms	12 ms	67x
SYSTEM	30,756	23.1 s	91 ms	253x
SYSTEM (Win10 1709)	43,211	118.7 s	111 ms	1,068x
SOFTWARE	117,488	745.6 s	292 ms	2,550x

cProfile shows why: in the Python backend, traversal time is dominated by per-record construct struct parsing and value decoding; with the Rust backend, the parser disappears from the profile entirely and the only remaining Python cost is constructing the returned Subkey dataclasses. The full profiles, per-hive digests and a disclaimer documenting the few intentional divergences (exception types on corrupted hives, cycle-guard behavior) are in regipy-rs/BENCHMARKS.md; python regipy-rs/benchmark.py regenerates the entire report.

The pure-Python parser remains the default and is unaffected when the Rust backend is not installed.

CLI

Parse the header

regipy-parse-header ~/Documents/TestEvidence/Registry/SYSTEM

Example output:

╒════════════════════════╤══════════╕
│ signature              │ b'regf'  │
├────────────────────────┼──────────┤
│ primary_sequence_num   │ 11639    │
├────────────────────────┼──────────┤
│ secondary_sequence_num │ 11638    │
├────────────────────────┼──────────┤
│ last_modification_time │ 0        │
├────────────────────────┼──────────┤
│ major_version          │ 1        │
├────────────────────────┼──────────┤
│ minor_version          │ 5        │
├────────────────────────┼──────────┤
│ file_type              │ 0        │
├────────────────────────┼──────────┤
│ file_format            │ 1        │
├────────────────────────┼──────────┤
│ root_key_offset        │ 32       │
├────────────────────────┼──────────┤
│ hive_bins_data_size    │ 10534912 │
├────────────────────────┼──────────┤
│ clustering_factor      │ 1        │
├────────────────────────┼──────────┤
│ file_name              │ SYSTEM   │
├────────────────────────┼──────────┤
│ checksum               │ 0        │
╘════════════════════════╧══════════╛
[2019-02-09 13:46:12.111654] WARNING: regipy.cli: Hive is not clean! You should apply transaction logs

When parsing the header of a hive, also checksum validation and transaction validations are done

Dump entire hive to disk (this might take some time)

regipy-dump ~/Documents/TestEvidence/Registry/NTUSER-CCLEANER.DAT -o /tmp/output.json

regipy-dump util can also output a timeline instead of a JSON, by adding the -t flag

Run relevant plugins on Hive

regipy-plugins-run ~/Documents/TestEvidence/Registry/SYSTEM -o /tmp/plugins_output.json

The hive type will be detected automatically and the relevant plugins will be executed. See the plugins section for more information

Compare registry hives

Compare registry hives of the same type and output to CSV (if -o is not specified output will be printed to screen)

regipy-diff NTUSER.dat NTUSER_modified.dat -o /tmp/diff.csv

Example output:

[2019-02-11 19:49:18.824245] INFO: regipy.cli: Comparing NTUSER.DAT vs NTUSER_modified.DAT
╒══════════════╤══════════════╤════════════════════════════════════════════════════════════════════════════════╤════════════════════════════════════════════════╕
│ difference   │ first_hive   │ second_hive                                                                    │ description                                    │
╞══════════════╪══════════════╪════════════════════════════════════════════════════════════════════════════════╪════════════════════════════════════════════════╡
│ new_subkey   │              │ 2019-02-11T19:46:31.832134+00:00                                               │ \Software\Microsoft\legitimate_subkey          │
├──────────────┼──────────────┼────────────────────────────────────────────────────────────────────────────────┼────────────────────────────────────────────────┤
│ new_value    │              │ not_a_malware: c:\temp\legitimate_binary.exe @ 2019-02-11 19:45:25.516346+00:00 │ \Software\Microsoft\Windows\CurrentVersion\Run │
╘══════════════╧══════════════╧════════════════════════════════════════════════════════════════════════════════╧════════════════════════════════════════════════╛
[2019-02-11 19:49:18.825328] INFO: regipy.cli: Detected 2 differences

Recover a registry hive, using transaction logs

regipy-process-transaction-logs NTUSER.DAT -p ntuser.dat.log1 -s ntuser.dat.log2 -o recovered_NTUSER.dat

After recovering, compare the hives with registry-diff to see what changed

Using as a library

Initiate the registry hive object

from regipy.registry import RegistryHive

reg = RegistryHive("/Users/martinkorman/Documents/TestEvidence/Registry/Vibranium-NTUSER.DAT")

Iterate recursively over the entire hive, from root key

for entry in reg.recurse_subkeys(as_json=True):
    print(entry)

regipy v6.3.0