blackboxprotobuf 1.0.1

Library for reading protobuf buffers without .proto definitions

BlackBox Protobuf Library

Note: This is a fork of the library found here. This original was written for adding protobuf reading to burp, this version strips out all burp related code and dependencies, and works with python3.

Description

Blackbox protobuf library is a Python module for decoding and re-encoding protobuf messages without access to the source protobuf descriptor file. This library provides a simple Python interface to encode/decode messages that can be integrated into other tools.

This library is targeted towards use in DFIR investigations where being able to read the content messages is critical and a protocol buffer definition may not be readily available.

Background

Protocol Buffers (protobufs) are a standard published by Google with accompanying libraries for binary serialization of data. Protocol buffers are defined by a .proto file known to both the sender and the receiver. The actual binary message does not contain information such as field names or most type information.

For each field, the serialized protocol buffer includes two pieces of metadata, a field number and the wire type. The wire type tells a parser how to parse the length of the field, so that it can be skipped if it is not known (one protocol buffer design goal is being able to handle messages with unknown fields). A single wire-type generally encompasses multiple protocol buffer types, for example the length delimited wire-type can be used for string, bytestring, inner message or packed repeated fields. See https://developers.google.com/protocol-buffers/docs/encoding#structure for the breakdown of wire types.

The protocol buffer compiler (protoc) does support a similar method of decoding protocol buffers without the definition with the --decode_raw option. However, it does not provide any functionality to re-encode the decoded message.

How it works

The library makes a best effort guess of the type based on the provided wire type (and occasionally field content) and builds a type definition that can be used to re-encode the data. In general, most fields of interest are likely to be parsed into a usable form. Users can optionally pass in custom type definitions that override the guessed type. Custom type definitions also allow naming of fields to improve user friendliness.

Future Work

• Allow import and export of type definitions to protobuf definition files.

Usage

Installation

pip install blackboxprotobuf

Interface

The main blackboxprotobuf module defines five functions, the core encoding/decoding functions, two convenience functions that encode/decode JSON strings and a function to validate type definition changes.

Decode

Decoding functions takes a protobuf bytestring, and optionally either a type definition or a known message name mapped to a type definition (in blackboxprotobuf.known_messages). If a type definition isn't provided, an empty message type is assumed and all types are derived from the protobuf binary.

The decoder returns a tuple containing a dictionary with the decoded data and a dictionary containing the generated type definition. If the input type definition does not include types for all fields in the message, the output type definitions will include type guesses for those fields.

Example use:

import blackboxprotobuf
import base64

data = base64.b64decode('KglNb2RpZnkgTWU=')
message,typedef = blackboxprotobuf.protobuf_to_json(data)
print(message)
print(typedef)

Encode

The encoding functions takes a Python dictionary containing the data and a type definition. Unlike decoding, the type definition is required and will fail if any fields are not defined. Generally, the type definition should be the output from the decoding function or a modified version thereof.

Example use:

import blackboxprotobuf
import base64

data = base64.b64decode('KglNb2RpZnkgTWU=')
message,typedef = blackboxprotobuf.decode_message(data)

message['5'] = 'Modified Me'

new_data = bytes(blackboxprotobuf.encode_message(message,typedef))
print(data)
print(new_data)

Type definition structure

The type definition object is a Python dictionary representing the type structure of a message, it includes a type for each field and optionally a name. Each entry in the dictionary represents a field in the message. The key should be the field number and the value is a dictionary containing attributes.

At the minimum the dictionary should contain the 'type' entry which contains a string identifier for the type. Valid type identifiers can be found in blackboxprotobuf/lib/types/type_maps.py.

Message fields will also contain one of two entries, 'message_typedef' or 'message_type_name'. 'message_typedef' should contain a second type definition structure for the inner message. 'message_type_name' should contain the string identifier for a message type previously stored in blackboxprotobuf.known_messages. If both are specified, the 'message_type_name' will be ignored.

Type Breakdown

The following is a quick breakdown of wire types and default values. See https://developers.google.com/protocol-buffers/docs/encoding for more detailed information from Google.

Variable Length Integers (varint)

The varint wire type represents integers with multiple bytes where one bit of each is dedicated to indicating if it is the last byte. This can be used to represent integers (signed/unsigned), boolean values or enums. Integers can be encoded using three variations:

• uint: Varint encoding with no representation of negative numbers.
• int: Standard encoding but inefficient for negative numbers (always 10 bytes).
• sint: Uses ZigZag encoding to efficiently represent negative numbers by mapping negative numbers into the integer space. For example -1 is converted to 1, 1 to 2, -2 to 3, and so on. This can result in drastically different numbers if a type is misinterpreted and either the original or incorrect type is sint.

The default is currently int with no ZigZag encoding.

Fixed32/64

The fixed length wire types have an implicit size based on the wire type. These support either fixed size integers (signed/unsigned) or fixed size floating point numbers (float/double). The default type for these is the floating point type as most integers are more likely to be represented by a varint.

Length Delimited

Length delimited wire types are prefixed with a varint indicating the length. This is used for strings, bytestrings, inner messages and packed repeated fields. Messages can generally be identified by validating if it is a valid protobuf binary. If it is not a message, the default type is a string/byte which are relatively interchangeable in Python.

Packed repeated fields are arrays of either varints or a fixed length wire type. Non-packed repeated fields use a separate tag (wire type + field number) for each element, allowing them to be easily identified and parsed. However, packed repeated fields only have the initial length delimited wire type tag. The parser is assumed to know the full type already for parsing out the individual elements. This makes this field type difficult to differentiate from an arbitrary byte string and will require user intervention to identify. In protobuf version 2, repeated fields had to be explicitly declared packed in the definition. In protobuf version 3, repeated fields are packed by default and are likely to become more common.