Tainted Digital Evidence and Privacy Protection in Blockchain-Based Systems

Full text

DFRWS 2020 EU eProceedings of the Seventh Annual DFRWS Europe
Tainted Digital Evidence and Privacy Protection in Blockchain-Based Systems
David Billard. HES-SO, University of Applied Sciences in Geneva, Switzerland
E-mail address: David.Billard@hesge.ch
Extended abstract
This paper focuses on an often-forgotten aspect of digital evidence
handling, when a court dismisses an evidence from a trial. Multiple rea-
sons can lead to dismiss an evidence: it can be challenged by a party during
an investigation or in front of the court, or simply dropped by the
prosecutor.
Of course, different countries apply different laws, but let's take a simple
example, that is quite universal. Bob is suspected to hold illegal child
pornography material. A warrant is issued and a police search is conducted
at Bob's house. During the search, a hard drive is seized and following the
police procedure, the hard drive is registered and a chain of custody is
initiated. Since this police body is a modern one, the hard drive is also
registered into the blockchain-based evidence inventory software.
Digital forensics experts examine the drive and find connections with
Alice, who seems deeply involved in child pornography. A police search is
therefore triggered on Alice and a USB stick with a lot of inculpatory evi-
dence is found at Alice's home. As required by the procedure, the USB stick
is registered into the same blockchain-based software.
Much later in the investigation, a defense lawyer raises the legality of the
first police search on serious grounds. The court follows the motion and
the first police search is dismissed. Since the second police search is a
direct offspring of the first, it is also dismissed from the case.
Now let's examine the same case, when a blockchain is used to implement
a chain of custody [1], [2]. The digital evidence found at Alice's and Bob's
are stored in the ledger, but there is no mechanism to delete a digital
evidence from the ledger, since it is precisely built to forbid any alteration,
deletion or cancellation.
In the absence of such mechanism, having this unique blockchain structure
available, there are at least two possible options in order to dismiss
transactions.
The first one is to delete the whole blockchain and to issue a new block-
chain, without the dismissed evidence. In practice, it means to start from
the root block and re-issue all the subsequent transactions (excepted the
transactions linked to the dismissed evidence of course). Although it is
theoretically doable, it means a huge effort of transaction and block vali-
dation, involving voting algorithms, and keeping track of all the blockchain
intra references. The reader can already notice that the computational
complexity is quite significant.
The second option is to issue undo-transactions whose purpose is to
indicate that the referenced transaction is void and cannot be used
anymore. It means that the blockchain contains two categories of trans-
actions: (1) the transactions for registering evidence and (2) the undo-
transactions for dismissing evidence.
This technique of using undo-transactions is widely used, since a long
time, in DataBase Management Systems (DBMS) for recovery or rollback
purposes[3]. Unfortunately, while it is well suited for DBMS, it brings some
issues in blockchain-based systems.
The major issue concerns the verification of the transaction validation. In
order for a user to check if a transaction is valid, the user will have to verify
if the chain of hashes and signatures has not be en broken since a particular
point in time (usually the begin of the blockchain). This check means that
the transaction has been correctly entered into the system and has been
validated following the rules.
But this check does not prove that the transaction is valid from a legal point
of view: the evidence linked to the transaction may have been dismissed
later. Therefore, the check process must continue until either: (1) it finds
the undo-transaction, then the transaction is not legally valid or (2) it
reaches the end of the blockchain, then the transaction is legally valid. The
reader will notice that the computational complexity of this check is
significantly higher than the single transaction verification protocol usu-
ally observed in blockchain.
Both solutions are unsatisfactory and we devised a solution where an
additional layer of software, AccessTX, provides access to the blockchain.
Let's name the chain of custody blockchain InventoryTX. The additional
layer, controlling access to InventoryTX, will first check if a transaction is
either legally dismissed or valid, before granting access. After the check is
done, the transaction will be validated through the normal processing of
the blockchain InventoryTX.
In order to check the legal validity of a transaction, the access control will
also use a blockchain, named InvalidatedTX. The payload of every trans-
action in InvalidatedTX contains the transaction ID related to a tainted
evidence. It is desirable that each transaction in InvalidatedTX is signed by
the jurisdiction issuing the removal of the tainted evidence.
The validation of each invalidating transaction is processed as in a normal
blockchain, since the root of InvalidatedTX. Only the nature of the inva-
lidating transaction differentiates it from a traditional blockchain.
An example might be the best way to illustrate the different components of
the proposed solution. We suppose the police searches Ms Marple's home.
This woman is suspected to host a suspected man running from the police.
Three evidence items are found at her home:
Agent Poirot found a USB key with the searched man identity doc-
uments and 1000 bitcoins;
Agent Ness found a notebook with pornographic contents and a
hyperlink to a web server;
Agent Loch found a love letter from the suspected man to Ms
Marple.
Later, the web site is investigated by agent Chris and it contains drug
recipes. The InventoryTX blockchain is built and has the look of Fig. 1.
The reader will notice that it is a generic representation of a blockchain
and that different authors in the literature may have additional fea-
tures.
Contents lists available at ScienceDirect
Forensic Science International: Digital Investigation
journal homepage: www.elsevier.com/locate/fsidi
https://doi.org/10.1016/j.fsidi.2020.300911
Forensic Science International: Digital Investigation 32 (2020) 300911

We suppose that, in our fictious example, the defense argues that porno-
graphic materials and drug recipes are not the subject of the search and
should be dismissed. The court follows this request and judges Roy and
Prince update the InvalidatedTX blockchain which is depicted in Fig. 2.
When parties will access the evidence stored into the InventoryTX, the
system will first look up in the InvalidatedTX to verify if the transaction
concerning the evidence is legally sound. Three scenarii are then possible:
If the transaction hash is absent from InvalidatedTX, and present in
InventoryTX then the system will serve the transaction payload,
which is usually a reference to a safe storage entity holding the
evidence content, or description.
If the transaction hash is absent from InvalidatedTX, and also absent
from InventoryTX then the system will raise a “Transaction not
found”exception.
If the transaction hash is present in InvalidatedTX then the system
will raise a “Transaction invalidated by court order #xxx”
exception.
This system possesses the advantage of being very lightweight. In the
absence of dismissed evidence, the cost for the lookup is in O(1), since
InvalidatedTX is empty. In the presence of dismissed evidence, the cost for
the lookup is in O(m) were mis the total number of dismissed evidence
records.
Our solution for dismissing tainted evidence do not erase the fact that the
evidence was once part of the procedure, but it will prevent the use of this
evidence by the parties. We believe that this algorithm will help in the
adoption of blockchain solutions by providing more flexibility in the evi-
dence management. Besides, this solution works with a majority of
blockchain implementation because it does not modify the blockchain
structure.
Furthermore, evidence data is separated from the blockchain transaction's
payload, that holds only metadata.
References
[1] A. H. Lone and R. N. Mir, “Forensic-chain: Blockchain based digital fo-
rensics chain of custody with PoC in Hyperledger Composer,”Digital
Investigation, vol. 28, pp. 44e55, Mar. 2019, https://doi.org/10.1016/j.diin.
2019.01.002.
[2] H. Al-Khateeb, G. Epiphaniou, and H. Daly, “Blockchain for Modern
Digital Forensics: The Chain-of-Custody as a Distributed Ledger,”in
Blockchain and Clinical Trial: Securing Patient Data, H. Jahankhani, S.
Kendzierskyj, A. Jamal, G. Epiphaniou, and H. Al-Khateeb, Eds. Cham:
Springer International Publishing, 2019, pp. 149e168.
[3] J. Gray and A. Reuter, Transaction Processing: Concepts and Techniques.
Morgan Kaufmann Publishers Inc., 1992.
Fig. 1. InventoryTX for the Marple case.Ă
Fig. 2. InvalidatedTX for the Marple case.Ă
Forensic Science International: Digital Investigation 32 (2020) 300911S2