Skip to content

mkorman90/regipy

Repository files navigation

regipy

OpenSSF Scorecard

⚠️ 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)

Iterate over a key and get all subkeys and their modification time:

for sk in reg.get_key('Software').iter_subkeys():
    print(sk.name, convert_wintime(sk.header.last_modified).isoformat())

Adobe 2019-02-03T22:05:32.525965
AppDataLow 2019-02-03T22:05:32.526047
McAfee 2019-02-03T22:05:32.526140
Microsoft 2019-02-03T22:05:32.526282
Netscape 2019-02-03T22:05:32.526352
ODBC 2019-02-03T22:05:32.526521
Policies 2019-02-03T22:05:32.526592

Get the values of a key:

reg.get_key('Software\Microsoft\Internet Explorer\BrowserEmulation').get_values(as_json=True)
[{'name': 'CVListTTL',
  'value': 0,
  'value_type': 'REG_DWORD',
  'is_corrupted': False},
 {'name': 'UnattendLoaded',
  'value': 0,
  'value_type': 'REG_DWORD',
  'is_corrupted': False},
 {'name': 'TLDUpdates',
  'value': 0,
  'value_type': 'REG_DWORD',
  'is_corrupted': False},
 {'name': 'CVListXMLVersionLow',
  'value': 2097211,
  'value_type': 'REG_DWORD',
  'is_corrupted': False},
 {'name': 'CVListXMLVersionHigh',
  'value': None,
  'value_type': 'REG_DWORD',
  'is_corrupted': False},
 {'name': 'CVListLastUpdateTime',
  'value': None,
  'value_type': 'REG_DWORD',
  'is_corrupted': False},
 {'name': 'IECompatVersionHigh',
  'value': None,
  'value_type': 'REG_DWORD',
  'is_corrupted': False},
 {'name': 'IECompatVersionLow',
  'value': 2097211,
  'value_type': 'REG_DWORD',
  'is_corrupted': False},
 {'name': 'StaleCompatCache',
  'value': 0,
  'value_type': 'REG_DWORD',
  'is_corrupted': False}]

Use as a plugin:

from regipy.plugins.ntuser.ntuser_persistence import NTUserPersistencePlugin
NTUserPersistencePlugin(reg, as_json=True).run()

{
	'Software\\Microsoft\\Windows\\CurrentVersion\\Run': {
		'timestamp': '2019-02-03T22:10:52.655462',
		'values': [{
			'name': 'Sidebar',
			'value': '%ProgramFiles%\\Windows Sidebar\\Sidebar.exe /autoRun',
			'value_type': 'REG_EXPAND_SZ',
			'is_corrupted': False
		}]
	}
}

Run all relevant plugins for a specific hive

from regipy.plugins.utils import run_relevant_plugins
reg = RegistryHive('/Users/martinkorman/Documents/TestEvidence/Registry/SYSTEM')
run_relevant_plugins(reg, as_json=True)

{
	'routes': {},
	'computer_name': [{
		'control_set': 'ControlSet001\\Control\\ComputerName\\ComputerName',
		'computer_name': 'DESKTOP-5EG84UG',
		'timestamp': '2019-02-03T22:19:28.853219'
	}]
}

Validation cases

Validation cases report

All new plugins should have one or more basic validation cases (which can be expanded in the future), for example:

from regipy.plugins.system.bam import BAMPlugin
from regipy_tests.validation.validation import ValidationCase


class NTUserUserAssistValidationCase(ValidationCase):
    # define your plugin class
    plugin = BAMPlugin
    # define the test file name, which should be present in `regipy_tests/data`
    test_hive_file_name = "SYSTEM_WIN_10_1709.xz"

    # Use `expected_entries` to test for presence of a few samples from the plugin results
    expected_entries = [
        {
            "sequence_number": 9,
            "version": 1,
            "sid": "S-1-5-90-0-1",
            "executable": "\\Device\\HarddiskVolume2\\Windows\\System32\\dwm.exe",
            "timestamp": "2020-04-19T09:09:35.731816+00:00",
            "key_path": "\\ControlSet001\\Services\\bam\\state\\UserSettings\\S-1-5-90-0-1",
        }
    ]

    # OR use `exact_expected_result` to test for an exact result:
    exact_expected_result = [
        {
            "sequence_number": 9,
            "version": 1,
            "sid": "S-1-5-90-0-1",
            "executable": "\\Device\\HarddiskVolume2\\Windows\\System32\\dwm.exe",
            "timestamp": "2020-04-19T09:09:35.731816+00:00",
            "key_path": "\\ControlSet001\\Services\\bam\\state\\UserSettings\\S-1-5-90-0-1",
        },
        {
            "sequence_number": 8,
            "version": 1,
            "sid": "S-1-5-90-0-1",
            "executable": "\\Device\\HarddiskVolume2\\Windows\\System32\\cmd.exe",
            "timestamp": "2020-04-19T09:09:34.544224+00:00",
            "key_path": "\\ControlSet001\\Services\\bam\\state\\UserSettings\\S-1-5-90-0-1",
        }
    ]

    expected_entries_count = 2

Development

Setting up for development

# Clone the repository
git clone https://github.com/mkorman90/regipy.git
cd regipy

# Install in development mode with all dependencies
pip install -e ".[full,dev]"

# Install pre-commit hooks
pre-commit install

Running tests

# Run all tests
pytest

# Run specific test files
pytest regipy_tests/tests.py
pytest regipy_tests/cli_tests.py

# Run plugin validation
PYTHONPATH=. python regipy_tests/validation/plugin_validation.py

Code quality

# Run linter
ruff check .

# Run formatter
ruff format .

# Run type checker
mypy regipy/

Testing GitHub Actions Locally

To test CI workflow changes locally before pushing, use act:

# Install act (Fedora)
sudo dnf install act-cli

# Install act (macOS)
brew install act

# Install act (other)
# See https://nektosact.com/installation/index.html

Make sure Docker is running, then:

# List available jobs
act -l

# Run the lint job
act -j lint

# Run all jobs for a push event
act push

# Run the test job with a specific Python version
act -j test

# Test the build job from publish workflow (simulates a release)
act release -j build --eventpath /dev/stdin <<< '{"action": "published"}'

Note: Some jobs may require secrets. You can provide them with:

act -j publish --secret PYPI_API_TOKEN=your_token

License

MIT

About

Regipy is an os independent python library for parsing offline registry hives

Resources

License

Security policy

Stars

278 stars

Watchers

5 watching

Forks

Packages

 
 
 

Contributors