febrezo / pymerkle

A Python library for constructing Merkle Trees and validating Log Proofs

pymerkle: A Python library for constructing Merkle Trees and validating Proofs

Complete documentation can be found at pymerkle.readthedocs.org

Pymerkle provides a class for binary balanced Merkle-trees (with possibly odd number of leaves) capable of generating audit-proofs, consistency-proofs and inclusion-tests. It supports all hash functions (including SHA3 variations) and encoding types, whereas defense against second-preimage attack is by default activated. It further provides flexible mechanisms for validating the generated proofs and thus easy verification of encrypted data.

It is a zero dependency library (with the inessential exception of tqdm for displaying progress bars).

Installation

pip3 install pymerkle --pre

Quick example

See also Usage and API

from pymerkle import MerkleTree, validateProof, validationReceipt

tree = MerkleTree()                                           # Empty SHA256/UTF-8 Merkle-tree with
                                                              # defense against second-preimage attack

for _ in range(665):                                          # Update the tree with 666 records
    tree.encryptRecord(bytes('%d-th record' % _, 'utf-8'))

_audit = tree.auditProof('12-th record')                      # Request audit-proof for the given record        

validateProof(target_hash=tree.rootHash(), proof=_audit)           # Quick validation of the above proof (True)

# Store the tree's current state for later use

oldhash   = tree.rootHash()
sublength = tree.length()

# Further encryption of files and objects

tree.encryptObject({'a': 0, 'b': 1})                          # One new leaf storing the provided
                                                              # object's digest
tree.encryptFileContent('../path/to/file')                    # One new leaf storing the digest
                                                              # of the provided file's content
tree.encryptFilePerLog('../logs/sample_log')                  # Many new leaves (one for each
                                                              # line of the provided file)

_consistency = tree.consistencyProof(oldhash, sublength)      # Request consistency-proof for the
                                                              # stored state of the Merkle-tree

_receipt = validationReceipt(
  target=tree.rootHash(),
  proof=_consistency)                                         # Validate proof with receipt                                            

Tree structure

Contrary to most implementations, the Merkle-tree is here always binary balanced, with all nodes except for the exterior ones (leaves) having two parents. This is achieved as follows: upon appending a block of new leaves, instead of promoting a lonely leaf to the next level or duplicating it, a bifurcation node gets created so that trees with the same number of leaves have always identical structure and input clashes among growing strategies be avoided. This standardization is further crucial for:

• fast generation of consistency-proof paths (based on additive decompositions in decreasing powers of 2)
• fast recalculation of the root-hash after appending a new leaf, since only the hashes at the tree's left-most branch need be recalculated
• memory efficiency, since the height as well as total number of nodes with respect to the tree's length is controlled to the minimum. For example, a tree with 9 leaves has 17 nodes in the present implementation, whereas the total number of nodes in the structure described here is 20.

This topology turns out to be identical with that of a binary Sekura tree, depicted in Section 5.4 of this paper. Follow the straightforward algorithm of the MerkleTree.update() method for further insight.

Encryption

Direct encryption of text (string, bytes, bytearray), JSON (dict) and files is supported. Use accordingly any of the following methods of the MerkleTree class (all of them internally invoking the .update() method for appending newly-created leaves): .encryptRecord(), .encryptFileConent(), .encryptFilePerLog(), .encryptObject(), .encryptObjectFromFile(), .encryptFilePerObject()

See API or Usage for details about arguments and precise functionality.

Proof validation

Direct validation of a Merkle-proof is performed using the validateProof() function, which modifies the status of the provided proof appropriately and returns the corresponding boolean. A more elaborate validation procedure includes generating a receipt with the validation result and storing at will the generated receipt as a .json file. This is achieved using the validationReceipt() function like in the above quick example.

See API or Usage for details about arguments and precise functionality.

Security

Defense against second-preimage attack

Defense against second-preimage attack is by default activated. Roughly speaking, it consists in the following security measures:

• Before calculating the hash of a leaf, prepend the corresponding record with the null hexadecimal 0x00

• Before calculating the hash of any interior node, prepend both of its parents' hashes with the unit hexadecimal 0x01

Read the tests/test_defense.py file inside the project's repository to see how to perform second-preimage attacks against the current implementation. In order to disable defense against second-preimage attack (say, for testing purposes), set security equal to False at construction:

tree = MerkleTree(security=False)

Deviation from bitcoin specification

In contrast to the bitcoin specification for Merkle-trees, lonely leaves are not duplicated in order for the tree to remain genuinely binary. Instead, creating bifurcation nodes at the rightmost branch allows the tree to remain both binary and balanced upon any update (see Tree structure above). As a consequence, even if strict security mode were deactivated (see above), the current implementation is structurally invulnerable to denial-of-service attacks exploiting the vilnerability described here (reported as CVE-2012-2459).

Persistence

On-disc persistence is not currently supported.

Given an instance of the MekleTree class, the minimum required information can be exported with the .export() method into a .json file; invoking the .loadFromFile() static method reloads the Merkle-tree from that file in its previously stored state. This can be useful for transmitting the tree's current state to a trusted party or retrieving the tree from a backup file. Reconstruction of the tree is uniquely determined by the sequence of stored hashes (see the section Tree structure to understand why).

See API or Usage for details about arguments and precise functionality.

Running tests

You need to have installed pytest. From inside the project's root directory type

pytest tests/

to run all tests. This might take up to 15-20 minutes, since crypto parts of the code are tested against all possible combinations of hash algorithm, encoding type and security mode (1620 combinations in total). Use the syntax and flag arguments of the pytest command to run only specific tests or have useful info about the tests printed.