A Quantitative Analysis of the Impact of Arbitrary Blockchain Content on Bitcoin Roman Matzutt1, Jens Hiller1, Martin Henze1, Jan Henrik Ziegeldorf1, Dirk Müllmann2, Oliver Hohlfeld1, and Klaus Wehrle1 1 Communication and Distributed Systems, RWTH Aachen University, Germany, {matzutt,hiller,henze,ziegeldorf,hohlfeld,wehrle}@comsys.rwth-aachen.de 2 Data Protection Research Institute, Goethe University, Frankfurt/Main, muellmann@jur.uni-frankfurt.de Abstract. Blockchains primarily enable credible accounting of digital events, e.g., money transfers in cryptocurrencies. However, beyond this original purpose, blockchains also irrevocably record arbitrary data, rang- ing from short messages to pictures. This does not come without risk for users as each participant has to locally replicate the complete blockchain, particularly including potentially harmful content. We provide the first systematic analysis of the benefits and threats of arbitrary blockchain content. Our analysis shows that certain content, e.g., illegal pornogra- phy, can render the mere possession of a blockchain illegal. Based on these insights, we conduct a thorough quantitative and qualitative anal- ysis of unintended content on Bitcoin’s blockchain. Although most data originates from benign extensions to Bitcoin’s protocol, our analysis re- veals more than 1600 files on the blockchain, over 99 % of which are texts or images. Among these files there is clearly objectionable content such as links to child pornography, which is distributed to all Bitcoin partic- ipants. With our analysis, we thus highlight the importance for future blockchain designs to address the possibility of unintended data insertion and protect blockchain users accordingly. 1 Introduction Bitcoin [45] was the first completely distributed digital currency and remains the most popular and widely accepted of its kind with a market price of ∼4750 USD per bitcoin as of August 31st, 2017 [14]. The enabler and key innovation of Bit- coin is the blockchain, a public append-only and tamper-proof log of all transac- tions ever issued. These properties establish trust in an otherwise trustless, com- pletely distributed environment, enabling a wide range of new applications, up to distributed general-purpose data management systems [69] and purely digital data-sharing markets [41]. In this work, we focus on the arbitrary, non-financial data on Bitcoin’s famous blockchain, which primarily stores financial transac- tions. This non-financial data fuels, e.g., digital notary services [50], secure re- leases of cryptographic commitments [16], or non-equivocation schemes [62]. However, since all Bitcoin participants maintain a complete local copy of the blockchain (e.g., to ensure correctness of blockchain updates and to bootstrap new users), these desired and vital features put all users at risk when objection- able content is irrevocably stored on the blockchain. This risk potential is exem- plified by the (mis)use of Bitcoin’s blockchain as an anonymous and irrevocable content store [40,56,35]. In this paper, we systematically analyse non-financial content on Bitcoin’s blockchain. While most of this content is harmless, there is also content to be considered objectionable in many jurisdictions, e.g., the depic- tion of nudity of a young woman or hundreds of links to child pornography. As a result, it could become illegal (or even already is today) to possess the block- chain, which is required to participate in Bitcoin. Hence, objectionable content can jeopardize the currently popular multi-billion dollar blockchain systems. These observations raise the question whether or not unintended content is ultimately beneficial or destructive for blockchain-based systems. To address this question, we provide the first comprehensive and systematic study of unin- tended content on Bitcoin’s blockchain. We first survey and explain methods to store arbitrary, non-financial content on Bitcoin’s blockchain and discuss poten- tial benefits as well as threats, most notably w.r.t. content considered illegal in different jurisdictions. Subsequently and in contrast to related work [56,40,12], we quantify and discuss unintended blockchain content w.r.t. the wide range of insertion methods. We believe that objectionable blockchain content is a pres- suring issue despite potential benefits and hope to stimulate research to mitigate the resulting risks for novel as well as existing systems such as Bitcoin. This paper is organized as follows. We survey methods to insert arbitrary data into Bitcoin’s blockchain in Section 2 and discuss their benefits and risks in Section 3. In Section 4, we systematically analyze non-financial content in Bitcoin’s blockchain and assess resulting consequences. We discuss related work in Section 5 and conclude this paper in Section 6. 2 Data Insertion Methods for Bitcoin Beyond intended recording of financial transactions, Bitcoin’s blockchain also allows for injection of non-financial data, either short messages via special trans- action types or even complete files by encoding arbitrary data as standard trans- actions. We first briefly introduce Bitcoin transactions and subsequently survey methods available to store arbitrary content on the blockchain via transactions. Bitcoin transactions transfer funds between a payer (sender) and a payee (receiver), who are identified by public-private key pairs. Payers announce their transactions to the Bitcoin network. The miners then publish these transactions in new blocks using their computational power in exchange for a fee. These fees vary, but averaged at 215 satoshi per Byte during August 2017 [4] (1 satoshi = 10−8 bitcoin). Each transaction consists of several input scripts, which unlock funds of previous transactions, and of several output scripts, which specify who receives these funds. To unlock funds, input scripts contain a signature for the previous transaction generated by the owner of the funds. To prevent malicious scripts from causing excessive transaction verification overheads, Bitcoin uses transaction script templates and expects peers to discard non-compliant scripts. Data Insertion Methods Input ScriptsOutput Scripts P2PK P2PKH P2SHP2MS P2SH Injectors SatoshiCryptoGraffiti Apertus StandardOP_RET. Non-St. Coinbase P2SHNon-St. Fig. 1: Bitcoin data insertion methods (italics show content insertion services) Method Payload Costs/B Eff. OP RET. 80 B 3.18–173.55 ct poor Coinbase 96 B — poor Non-St. Out. 99 044 B 1.03–198.05 ct poor Non-St. In. med. P2PK 85 345 B 1.24–207.79 ct high P2PKH 58 720 B 1.87–197.58 ct high P2MS 92 625 B 1.11–234.33 ct high P2SH Out. 62 400 B 1.77–195.54 ct high P2SH In. 99 018 B 1.03–225.61 ct high Table 1: Payload, costs, and efficiency of low-level data insertion methods Figure 1 shows the insertion methods for non-financial data we identified in Bitcoin. We distinguish low-level data insertion methods inserting small data chunks and content insertion services, which systematically utilize the low-level methods to insert larger chunks of data. In the following, we refer to non-financial blockchain data as content if it has a self-contained structure, e.g., a file or read- able text, or as data otherwise, e.g., fragments inserted via a low-level method. 2.1 Low-level Data Insertion Methods We first survey the efficiency of the low-level data insertion methods w.r.t. to in- sertable payload and costs per transaction (Table 1). To this end, we first explain our comparison methodology, before we detail i) intended data insertion meth- ods (OP RETURN and coinbase), ii) utilization of non-standard transactions, and iii) manipulation of standard transactions to insert arbitrary data. Comparison Methodology. We measure the payload per transaction (PpT), i.e., the number of non-financial Bytes that can be added to a single standard- sized transaction (≤ 100 000 B). Costs are given as the minimum and maximum costs per Byte (CpB) for the longest data chunk a transaction can hold, and for inserting 1 B. Costs are inflicted by paying transaction fees and possibly burning currency (at least 546 satoshi per output script), i.e., making it unspendable. For our cost analysis we assume Bitcoin’s market price of 4748.25 USD as of August 31st, 2017 [14] and the average fees of 215 satoshi per Byte as of August 2017 [4]. Note that high variation of market price and fees results in frequent changes of presented absolute costs per Byte. Finally, we rate the overall efficiency of an approach w.r.t. insertion of arbitrary-length content. Intuitively, a method is efficient if it allows for easy insertion of large payloads at low costs. OP RETURN. This special transaction template allows attaching one small data chunk to a transaction and thus provides a controlled channel to an- notate transactions without negative side effects. E.g., in typical implementa- tions peers increase performance by caching spendable transaction outputs and OP RETURN outputs can safely be excluded from this cache. However, data chunk sizes are limited to 80 B per transaction. Coinbase. In Bitcoin, each block contains exactly one coinbase transaction, which introduces new currency into the system to incentivize miners to dedi- cate their computational power to maintain the blockchain. The input script of coinbase transactions is up to 100 B long and consists of a variable-length field encoding the new block’s position in the blockchain [9]. Stating a larger size than the overall script length allows placing arbitrary data in the resulting gap. This method is inefficient as only active miners can insert only small data chunks. Non-standard Transactions. Transactions can deviate from the approved transaction templates [48] via their output scripts as well as input scripts. In the- ory, such transactions can carry arbitrarily encoded data chunks. Transactions using non-standard output scripts can carry up to 96.72 KiB at comparably low costs. However, they are inefficient as miners ignore them with high probability. Yet, non-standard output scripts occasionally enter the blockchain if miners in- sufficiently check them (cf. Section 4.2). Contrarily, non-standard input scripts are only required to match their respective output script. Hence, input scripts can be altered to carry arbitrary data if their semantics are not changed, e.g., by using dead conditional branches. This makes non-standard input scripts slightly better suited for large-scale content insertion than non-standard output scripts. Standard Financial Transactions. Even standard financial transactions can be (mis)used to insert data using mutable values of output scripts. There are four approved templates for standard financial transactions: Pay to public-key (P2PK) and pay to public-key hash (P2PKH) transactions send currency to a dedicated receiver, identified by an address derived from her private key, which is required to spend any funds received [48]. Similarly, multi-signature (P2MS) transactions require m out of n private keys to authorize payments. Pay to script hash (P2SH) transactions refer to a script instead of keys to enable complex spending conditions [48], e.g., to replace P2MS [10]. The respective public keys (P2PK, P2MS) and script hash values (P2PKH, P2SH) can be replaced with ar- bitrary data as Bitcoin peers can not verify their correctness before they are ref- erenced by a subsequent input script. While this method can store large amounts of content, it involves significant costs: In addition to transaction fees, the user must burn bitcoins as she replaces valid receiver identifiers with arbitrary data (i.e., invalid receiver identities), making the output unspendable. Using multi- ple outputs enables PpTs ranging from 57.34 KiB (P2PKH) to 96.70 KiB (P2SH inputs) at CpBs from 1.03 ct to 1.87 ct. As they behave similarly w.r.t. data in- sertion, we collectively refer to all standard financial transactions as P2X in the following. P2SH scripts also allow for efficient data insertion into input scripts as P2SH input scripts are published with their redeem script. Due to miners’ verification of P2SH transactions, transaction are not discarded if the redeem script is not template-compliant (but the overall P2SH transaction is). We now survey different services that systematically leverage the discussed data insertion methods to add larger amounts of content to the blockchain. 2.2 Content Insertion Services Content insertion services rely on the low-level data insertion methods to add content, i.e., files such as documents or images, to the blockchain. We identify four conceptually different content insertion services and present their protocols. CryptoGraffiti. This web-based service [30] reads and writes messages and files from and to Bitcoin’s blockchain. It adds content via multiple P2PKH output scripts within a single transaction, storing up to 60 KiB of content. To retrieve previously added content, CryptoGraffiti scans for transactions that either con- sist of at least 90 % printable characters or contain an image file. Satoshi Uploader. The Satoshi Uploader [56] inserts content using a single transaction with multiple P2X outputs. The inserted data is stored together with a length field and a CRC32 checksum to ease decoding of the content. P2SH Injectors. Several services [35] insert content via slightly varying P2SH input scripts. They store chunks of a file in P2SH input scripts. To ensure file integrity, the P2SH redeem scripts contain and verify hash values of each chunk. Apertus. This service [29] allows fragmenting content over multiple transac- tions using an arbitrary number of P2PKH output scripts. Subsequently, these fragments are referenced in an archive stored on the blockchain, which is used to retrieve and reassemble the fragments. The chosen encoding optionally allows augmenting content with a comment, file name, or digital signature. To conclude, Bitcoin offers various options to insert arbitrary, non-financial data. These options range from small-scale data insertion methods exclusive to active miners to services that allow any user to store files of arbitrary length. This wide spectrum of options for data insertion raises the question which benefits and risks arise from storing content on Bitcoin’s blockchain. 3 Benefits and Risks of Arbitrary Blockchain Content Bitcoin’s design includes several methods to insert arbitrary, non-financial data into its blockchain in both intended and unintended ways. In this section, we discuss potential benefits of engraving arbitrary data into Bitcoin’s blockchain as well as risks of (mis)using these channels for content insertion. 3.1 Benefits of Arbitrary Blockchain Content Besides the manipulation of standard financial transactions, Bitcoin offers coin- base and OP RETURN transactions as explicit channels to irrevocably insert small chunks of non-financial data into its blockchain (cf. Section 2). As we discuss in the following, each insertion method has distinguishing benefits: OP RETURN. Augmenting transactions with short pieces of arbitrary data is beneficial for a wide area of applications [40,12,62]. Different services use OP RETURN to link non-financial assets, e.g., vouchers, to Bitcoin’s block- chain [40,12], to attest the existence of digital documents at a certain point of time as a digital notary service [58,50,12], to realize distributed digital rights management [70,12], or to create non-equivocation logs [62,8]. Coinbase. Coinbase transactions differ from OP RETURN as only miners, who dedicate significant computational resources to maintain the blockchain, can use them to add extra chunks of data to their newly mined blocks. Beyond advertisements or short text messages [40], coinbase transactions can aid the mining process. Adding random bytes to the coinbase transactions allows miners to increase entropy when repeatedly testing random nonces to solve the proof- of-work puzzle [48]. Furthermore, adding identifiable voting flags to transactions enables miners to vote on proposed features, e.g., the adoption of P2SH [10]. Large-scale Data Insertion. Engraving large amounts of data into the block- chain creates a long-term non-manipulable file storage. This enables, e.g., the archiving of historical data or censorship-resistant publication, which helps pro- tecting whistleblowers or critical journalists [66]. However, their content is repli- cated to all users, who do not have a choice to reject storing it. Hence, non-financial data on the blockchain enables new applications that leverage Bitcoin’s security guarantees. In the following, we discuss threats of forcing honest users to download copies of all blockchain content. 3.2 Risks of Arbitrary Blockchain Content Despite potential benefits of data in the blockchain, insertion of objectionable content can put all participants of the Bitcoin network at risk [43,11,40], as such unwanted content is unchangeable and locally replicated by each peer of the Bitcoin network as benign data. To underpin this threat, we first derive an extensive catalog of content that poses high risks if possessed by individuals and subsequently argue that objectionable blockchain content is able to harm honest users. In the following, we identify five categories of objectionable content: Copyright Violations. With the advent of file-sharing networks, pirated data has become a huge challenge for copyright holders. To tackle this problem, copy- right holders predominantly target users that actively distribute pirated data. E.g., German law firms sue users who distribute copyright-protected content via file-sharing networks for fines on behalf of the copyright holders [28]. In re- cent years, prosecutors also convicted downloaders of pirated data. For instance, France temporarily suspended users’ Internet access and subsequently switched to issuing high fines [36]. As users distribute their blockchain copy to new peers, copyright-protected material on the blockchain can thus provoke legal disputes about copyright infringement. Malware. Another threat is to download malware [20,42], which could poten- tially be spread via blockchains [31]. Malware has serious consequences as it can destroy sensitive documents, make devices inoperable, or cause financial losses [34]. Furthermore, blockchain malware can irritate users as it causes an- tivirus software to deny access to important blockchain files. E.g., Microsoft’s antivirus software detected a non-functional virus signature from 1987 on the blockchain, which had to be fixed manually [68]. Privacy Violations. By disclosing sensitive personal data, individuals can harm their own privacy and that of others. This threat peaks when individuals deliberately violate the privacy of others, e.g., by blackmailing victims under the threat of disclosing sensitive data about them on the blockchain. Real-world manifestations of these threats are well-known, e.g., non-consensually releasing private nude photos or videos [54] or fully disclosing an individual’s identity to the public with malicious intents [21]. Jurisdictions such as the whole European Union begin to actively prosecute the unauthorized disclosure and forwarding of private information in social networks to counter this novel threat [5]. Politically Sensitive Content. Governments have concerns regarding the leakage of classified information such as state secrets or information that other- wise harms national security, e.g., propaganda. Although whistleblowers reveal nuisances such as corruption, they force all blockchain users to keep a copy of leaked material. Depending on the jurisdiction, the intentional disclosure or the mere possession of such content may be illegal. While, e.g., the US government usually tends to prosecute intentional theft or disclosure of state secrets [63], in China the mere possession of state secrets can result in longtime prison sen- tences [49]. Furthermore, China’s definition of state secrets is vague [49] and covers, e.g., “activities for safeguarding state security” [60]. Such vague allega- tions w.r.t. state secrets have been applied to critical news in the past [18,24]. Illegal and Condemned Content. Some categories of content are virtually universally condemned and prosecuted. Most notably, possession of child pornog- raphy is illegal at least in the 112 countries [64] that ratified an optional protocol to the Convention on the Rights of the Child [65]. Religious content such as cer- tain symbols, prayers, or sacred texts can be objectionable in extremely religious countries that forbid other religions and under oppressive regimes that forbid re- ligion in general. As an example, possession of items associated with an objected religion, e.g., Bibles in Islamist countries, or blasphemy have proven risky and were sometimes even punished by death [13,38]. In conclusion, a wide range of objectionable content can cause direct harm if possessed by users. In contrast to systems such as social media platforms, file-sharing networks, or online storage systems, such content can be stored on blockchains anonymously and irrevocably. Since all blockchain data is down- loaded and persistently stored by users, they are liable for any objectionable content added to the blockchain by others. Consequently, it would be illegal to participate in a blockchain-based systems as soon as it contains illegal content. While this risk has previously been acknowledged [43], definitive answers re- quire court rulings yet to come. However, considering legal texts we anticipate a high potential for illegal blockchain content to jeopardize blockchain-based sys- tem such as Bitcoin in the future. Our belief stems from the fact that, w.r.t. child pornography as an extreme case of illegal content, legal texts from countries such as the USA [47], England [3], Ireland [32] deem all data illegal that can be con- verted into a visual representation of illegal content. As we stated in Section 2, it is easily possible to locate and reassemble such content on the blockchain. Hence, even though convertibility usually covers creating a visual representation by, e.g., decoding an image file, we expect that the term can be interpreted to include blockchain data in the future. For instance, this is already covered implicitly by German law, as a person is culpable for possession of illegal content if she knowingly possesses an accessible document holding said content [2]. It is criti- cal here that German law perceives the hard disk holding the blockchain as an document [1] and that users can easily reassemble any illegal content within the blockchain. Furthermore, users can be assumed to knowingly maintain control over such illegal content w.r.t. German law if sufficient media coverage causes the content’s existence to become public knowledge among Bitcoin users [61], as has been attempted by Interpol [31]. We thus believe that legislators will speak law w.r.t. non-financial blockchain content and that this has the potential to jeopardize systems such as Bitcoin if they hold illegal content. 4 Blockchain Content Landscape To understand the landscape of non-financial blockchain data and assess its potentials and risks, we thoroughly analyze Bitcoin’s blockchain as it is the most widely used blockchain today. Especially, we are interested in i) the degree of utilization of data and content insertion methods, ii) the temporal evolution of data insertion, and iii) the types of content on Bitcoin’s blockchain, especially w.r.t. objectionable content. In the following, we first outline our measurement methodology before we present an overview and the evolution of non-financial data on Bitcoin’s blockchain. Finally, we analyze files stored on the blockchain to derive if any objectionable content is already present on the blockchain. 4.1 Methodology We detect data-holding transactions recorded on Bitcoin’s blockchain based on our study of data insertion methods and content insertion services (cf. Section 2). We distinguish detectors for data insertion methods and detectors for content insertion services. To reduce false positives, e.g., due to public-key hash values that resemble text, we exclude all standard transaction outputs that include already-spent funds from analysis. This is sensible as data-holding transactions replace public keys or hashes such that spending requires computing correspond- ing private keys or pre-images, which is assumed to be infeasible. Contrarily, even though we thoroughly analyzed possible insertion methods, there is still a chance that we do not exhaustively detect all non-financial data. Nevertheless, our con- tent type analysis establishes a solid lower bound as we only consider readable files retrieved from Bitcoin’s blockchain. In the following, we explain the key characteristics of the two classes of our blockchain content detectors. Low-level Insertion Method Detectors. The first class of detectors is tai- lored to match individual transactions that are likely to contain non-financial data (cf. Section 2.1). These detectors detect manipulated financial transactions as well as OP RETURN, non-standard, and coinbase transactions. Our text detector scans for P2X output scripts for mutable values containing ≥ 90 % printable ASCII characters (to avoid false positives). The detector returns the concatenation of all output scripts of the same transaction that contain text. Finally, we consider all coinbase and OP RETURN transactions as well as non-standard output scripts. We detect coinbase transactions based on the length field mismatch described in Section 2.1. OP RETURN scripts are detectable as they always begin with an OP RETURN operation. Non-standard output scripts comprise all output scripts which are not template-conform. 2009 2011 2013 2015 2017 101 103 105 107 T ra n sa c ti o n s [# ] OP RET. Coinb. Non-St. P2X P2SH Input Fig. 2: Cumulative numbers of detected transactions per data insertion method 2013 2014 2015 2016 2017 0.0 0.2 0.4 0.6 0.8 1.0 1.2 P re se n c e in T X s [% ] OP RET. P2X P2SH Input Fig. 3: Ratio of transactions that utilize data insertion methods Service Detectors. We implemented detectors specific to the content insertion services we identified in Section 2.2. These service-specific detectors enable us to detect and extract files based on the services’ protocols. These detectors also track the data insertion method used in service-created transactions. The CryptoGraffiti detector matches transactions with an output that sends a tip to a public-key hash controlled by its provider. For such a transaction, we concatenate all mutable values of output scripts that spend fewer than 10 000 satoshi and store them in a file. This threshold is used to ignore non- manipulated output scripts, e.g., the service provider spending their earnings. To detect a Satoshi Uploader transaction, we concatenate all of its mutable values that spend the same small amount of bitcoins. If we find the first eight bytes to contain a valid combination of length and CRC32 checksum for the transaction’s payload, we store the payload as an individual file. We detect P2SH Injector content based on redeem scripts containing more than one hash operation (standard transactions use at most one). We then ex- tract the concatenation of the second inputs of all redeem scripts (the first one contains a signature) of a transaction as one file. Finally, the Apertus detector recursively scans the blockchain for Apertus archives, i.e., Apertus-encoded lists of previous transaction identifiers. Once a referred Apertus payload does not constitute another archive, we retrieve its payload file and optional comment by parsing the Apertus protocol. Suspicious Transaction Detector. To account for less wide-spread insertion services, we finally analyze standard transactions that likely carry non-financial data but are not detected otherwise. We only consider transactions with at least 50 suspicious outputs, i.e., roughly 1 KiB of content. We consider a set of outputs suspicious if all outputs i) spend the same small amount (< 10 000 satoshi) and ii) are unspent. This detector trades off detection rate against false-positive rate. Due to overlaps with service detectors, we exclude matches of this detector from our quantitative analysis, but discuss individual findings in Section 4.3. 4.2 Utilization of Data Insertion Methods Data and content insertion in Bitcoin has evolved over time, transitioning from single miners exploiting coinbase transactions to sophisticated services that en- able the insertion of whole files into the blockchain. We study this evolution in 645 In se rt io n s/ M o n th [# ] 2013 2014 2015 2016 2017 0 50 100 150 P2SH Injectors CryptoGraffiti Satoshi Uploader Apertus Fig. 4: Number of files inserted via con- tent insertion services per month 2013 2014 2015 2016 2017 0 2 4 6 8 T X si z e s [M iB ] P2SH Injectors CryptoGraffiti Satoshi Uploader Apertus Fig. 5: Cumulative sizes of transactions from content insertion services terms of used data insertion methods as well as content insertion services and quantify the amount of blockchain data using our developed detectors. Our key insights are that OP RETURN constitutes a well-accepted success story while content insertion services are currently only infrequently utilized. However, the introduction of OP RETURN did not shut down other insertion methods, e.g., P2X manipulation, which enable single users to insert objectionable content. Our measurements are based on Bitcoin’s complete blockchain as of August 31st, 2017, containing 482 870 blocks and 250 845 217 transactions with a total disk size of 122.64 GiB. We first analyze the popularity of different data inser- tion methods and subsequently turn towards the utilization of content insertion services to assess how non-financial data enters the blockchain. Data Insertion Methods. As described in Section 2.1, OP RETURN and coinbase transactions constitute intended data insertion methods, whereas P2X and non-standard P2SH inputs manipulate legitimate transaction templates to contain arbitrary data. Figure 2 shows the cumulative number of transactions containing non-financial data on a logarithmic scale. In total, our detectors found 3 535 855 transactions carrying a total payload of 118.53 MiB, i.e., only 1.4 % of Bitcoin transactions contain non-financial data. However, we strive to further un- derstand the characteristics of non-financial blockchain content as even a single instance of objectionable content can potentially jeopardize the overall system. The vast majority of extracted transactions are OP RETURN (86.8 % of all matches) and coinbase (13.13 %) transactions. Combined, they constitute 95.90 MiB (80.91 % of all extracted data). Out of all blocks, 96.15 % have content- holding coinbase transactions. While only 0.26 % of these contain ≥ 90 % print- able text, 33.49 % of them contain ≥ 15 consecutive printable ASCII characters (mostly surrounded by data without obvious structure). Of these short messages, 14.39 % contain voting flags for new features (cf. Section 3.1). Apart from this, miners often advertise themselves or leave short messages, e.g., prayer verses. OP RETURN transactions were introduced in 2013 to offer a benign way to augment single transactions with non-financial data. This feature is widely used, as shown by Figure 3. Among all methods, OP RETURN is the only one to be present with a rising tendency, with currently 1.2 % of all transactions containing OP RETURN outputs. These transactions predominantly manage off-blockchain assets or originate from notary services [12]. While P2X transactions are contin- uously being manipulated, they make up only 0.02 % of all transactions; P2SH inputs are virtually irrelevant. Hence, short non-financial data chunks are well- accepted, viable extensions to the Bitcoin system (cf. Section 3.1). P2X transactions are asymmetric w.r.t. the number and sizes of data-carrying transactions. Although constituting only 1.6 % of all detector hits, they make up 9.08 % of non-financial data (10.76 MiB). This again highlights the high content- insertion efficiency of P2X transactions (cf. Section 2.1). Finally, we discuss non-standard transactions and non-standard P2SH in- put scripts. In total, we found 1703 transactions containing non-standard out- puts. The three first non-standard transactions (July 2010) repeatedly used the OP CHECKSIG operation. We dedicate this to an attempted DoS attack that tar- gets to cause high verification times. Furthermore, we found 23 P2PKH transac- tions from October 2011 that contained OP 0 instead of a hash value. The steady increase of non-standard transactions in 2012 is due to scripts that consist of 32 seemingly random bytes. Contrarily, P2SH input scripts sporadically carry non- standard redeem scripts and are then often used to insert larger data chunks (as they are used by P2SH Injectors). This is due to P2SH scripts not being checked for template conformity. We found 888 such transactions holding 8.37 MiB of data. Although peers should reject such transactions [48], they still often man- age to enter the blockchain. Non-standard P2SH scripts even carry a substantial amount of data (7.07 % of the total data originate from P2SH Injectors). Content Insertion Services. We now investigate to which extent content insertion services are used to store content on Bitcoin’s blockchain. Figure 4 shows utilization patterns for each service and Figure 5 shows the cumulative size of non-financial data inserted via the respective service. Notably, only few users are likely responsible for the majority of service-inserted content. In total, content insertion services account for 16.12 MiB of non-financial data. More than a half of this content (8.37 MiB) originates from P2SH In- jectors. The remainder was mostly inserted using Apertus (21.70 % of service- inserted data) and Satoshi Uploader (21.24 %). Finally, CryptoGraffiti accounts for 0.82 MiB (5.10 %) of content related to content insertion services. In the following, we study how the individual services have been used over time. Our key observation is that both CryptoGraffiti and P2SH Injectors are in- frequently but steadily used; since 2016 we recognize on average 23.65 data items being added per month using these services. Contrarily, Apertus has been used only 26 times since 2016, while the Satoshi Uploader has not been used at all. In fact, the Satoshi Uploader was effectively used only during a brief period: 92.73 % of all transactions emerged in April 2013. During this time, the service was used to upload four archives, six backup text files, and a PDF file. Although Apertus and the Satoshi Uploader have been used only infrequently, together they constitute 64.32 % of all P2X data we detected. This stems from the utilization of those services to engrave files into the blockchain, e.g., archives or documents (Satoshi Uploader), or images (Apertus). Similarly, P2SH Injectors are used to backup conversations regarding development of the Bitcoin client, especially online chat logs, forum threads, and emails, with a significant peak File Via Service? Overall File Via Service? Overall Type yes no Portion Type yes no Portion Text 1353 54 87.07 % Archive 4 0 0.25 % Images 144 2 9.03 % Audio 2 0 0.12 % HTML 45 0 2.78 % PDF 2 0 0.12 % Source Code 7 3 0.62 % Total 1557 59 100.0 % Table 2: Distribution of blockchain file types according to our content-insertion- service and suspicious-transactions detectors. utilization between May and June 2015 (76.46 % of P2SH Injector matches). Es- pecially Apertus is well-suited for this task as files are spread over multiple trans- actions. Based on the median, the average Apertus file has a size of 17.15 KiB and is spread over 10 transactions, including all overheads. The largest Aper- tus file is 310.72 KiB large (including overheads), i.e., three times the size of a standard transaction, and is spread over 96 transactions. The most heavily frag- mented Apertus file is even spread over 664 transactions. Contrarily, 95.7 % of CryptoGraffiti matches are short text messages with a median length of 80 Byte. In conclusion, content insertion services are only infrequently used with vary- ing intentions and large portions of content was uploaded in bursts, indicating that only few users are likely responsible for the majority of service-inserted blockchain content. While CryptoGraffiti is mostly used to insert short text messages that also fit into one OP RETURN transaction, other services are pre- dominantly used to store, e.g., images or documents. As such files can constitute objectionable content, we further investigate them in the following. 4.3 Investigating Blockchain Files After quantifying basic content insertion in Bitcoin, we now focus on readable files that are extractable from the blockchain. We refer to files as findings of our content-insertion-service or suspicious-transaction detectors that are viewable using appropriate standard software. We reassemble fragmented files only if this is unambiguously possible, e.g., via an Apertus archive. Out of the 22.63 MiB of blockchain data not originating from coinbase or OP RETURN transactions, we can extract and analyze 1557 files with meaningful content. In addition to these, we could extract 59 files using our suspicious-transaction detector (92.25 % text). Table 2 summarizes the different file types of the analyzed files. The vast majority are text-based files and images (99.34 %). In the following, we discuss our findings with respect to objectionable con- tent. We manually evaluated all readable files with respect to the problematic categories we identified in Section 3.2. This analysis reveals that content from all those categories already exists in Bitcoin’s blockchain today. For each of these categories, we discuss the most severe examples. To protect the safety and pri- vacy of individuals, we omit personal identifiable information and refrain from providing exact information on the location of critical content in the blockchain. Copyright Violations. We found seven files that publish (intellectual) property and showcase Bitcoin’s potential to aid copyright violations. Engraved are the text of a book, a copy of the original Bitcoin paper [45,56], and two short textual white papers. Furthermore, we found two leaked cryptographic keys: one RSA private key and a firmware secret key. Finally, the blockchain contains a so-called illegal prime, encoding software to break the copy protection of DVDs [56]. Malware. We could not find actual malware in Bitcoin’s blockchain. How- ever, an individual non-standard transaction contains a non-malicious cross-site scripting detector. A security researcher inserted this small piece of code which, if interpreted by an online blockchain parser, notifies the author about the vul- nerability. Such malicious code could become a threat for users as most websites offering an online blockchain parser also offer online Bitcoin accounts. Privacy Violations. Users store memorable private moments on the block- chain. We extracted six wedding-related images and one image showing a group of people, labeled with their online pseudonyms. Furthermore, 609 transactions contain online public chat logs, emails, and forum posts discussing Bitcoin, in- cluding topics such as money laundering. Storing private chat logs on the block- chain can, e.g., leak single user’s private information irrevocably. Moreover, third parties can release information without knowledge nor consent of affected users. Most notably, we found at least two instances of doxing, i.e., the complete dis- closure of another individual’s personal information. This data includes phone numbers, addresses, bank accounts, passwords, and multiple online identities. Recently, jurisdictions such as the European Union began to punish such serious privacy violations, including the distribution of doxing data [5]. Again, carrying out such assaults via blockchains fortifies the problem due to their immutability. Politically Sensitive Content. The blockchain has been used by whistleblow- ers as a censorship-resistant permanent storage for leaked information. We found backups of the WikiLeaks Cablegate data [37] as well as an online news arti- cle concerning pro-democracy demonstrations in Hong Kong in 2014 [25]. As stated in Section 3.2, restrictive governments are known to prosecute the pos- session of such content. For example, state-critical media coverage has already put individuals in China [18] or Turkey [24] at the risk of prosecution. Illegal and Condemned Content. Bitcoin’s blockchain contains at least eight files with sexual content. While five files only show, describe, or link to mildly pornographic content, we consider the remaining three instances objectionable for almost all jurisdictions: Two of them are backups of link lists to child pornog- raphy, containing 274 links to websites, 142 of which refer to Tor hidden services. The remaining instance is an image depicting mild nudity of a young woman. In an online forum this image is claimed to show child pornography, albeit this claim cannot be verified (due to ethical concerns we refrain from providing a ci- tation). Notably, two of the explicit images were only detected by our suspicious- transaction detector, i.e., they were not inserted via known services. While largely harmless, potentially objectionable blockchain content is infre- quently inserted, e.g., links to alleged child pornography or privacy violations. We thus believe that future blockchain designs must proactively cope with objec- tionable content. Peers can, e.g., filter incoming transactions or revert content- holding transactions [11,51], but this must be scalable and transparent. 5 Related Work Previous work related to ours comprises i) mitigating the distribution of objec- tionable content in file-sharing peer-to-peer networks, ii) studies on Bitcoin’s blockchain, iii) reports on Bitcoin’s susceptibility for content insertion, and iv) approaches to retrospectively remove blockchain content. The trade-off between enabling open systems for data distribution and risking that unwanted or even illegal content is being shared is already known from peer-to-peer networks. Peer-to-peer-based file-sharing protocols typically limit the spreading of objectionable public content by tracking the reputation of users offering files [6,26,55,73] or assigning a reputation to files themselves [19,67]. This way, users can reject objectionable content or content from untrustworthy sources. Contrarily, distributed content stores usually resort to encrypt private files before outsourcing them to other peers [17,7]. By storing only encrypted files, users can plausibly deny possessing any content of others and can thus obliviously store it on their hard disk. Unfortunately, these protection mechanisms are not applicable to blockchains, as content cannot be deleted once it has been added to the blockchain and the utilization of encryption cannot be enforced reliably. Bitcoin’s blockchain was analyzed w.r.t. different aspects by numerous stud- ies. In a first step, multiple research groups [53,33,71,72,39] studied the currency flows in Bitcoin, e.g., to perform wealth analyses. From a different line of re- search, several approaches focused on user privacy and investigated the identities used in Bitcoin [52,46,44,59,23]. These works analyzed to which extent users can be de-anonymized by clustering identities [52,46,44,59,23] and augmenting these clusters with side-channel information [52,44,59,23]. Finally, the blockchain was analyzed w.r.t. the use cases of OP RETURN transactions [12]. While this work is very close to ours, we provide a first comprehensive study of the complete landscape of non-financial data on Bitcoin’s blockchain. The seriousness of objectionable content stored on public blockchains has been motivated by multiple works [56,57,43,11,40,51]. These works, however, fo- cus on reporting individual incidents or consist of preliminary analyses of the distribution and general utilization of content insertion. To the best of our knowl- edge, this paper gives the first comprehensive analysis of this problem space, including a categorization of objectionable content and a survey of potential risks for users if such content enters the blockchain. In contrast to previously considered attacks on Bitcoin’s ecosystem [22,27], illegal content can be inserted instantly at comparably low costs and can put all participants at risk. The utilization of chameleon hash functions [15] to chain blocks recently opened up a potential approach to mitigate unwanted or illegal blockchain con- tent [11]. Here, a single blockchain maintainer or a small group of maintainers can retrospectively revert single transactions, e.g., due to illegal content. To overcome arising trust issues, µchain [51] leverages the consensus approach of traditional blockchains to vote on alterations of the blockchain history. As these approaches tackle unwanted content for newly designed blockchains, we seek to motivate a discussion on countermeasures also for existing systems, e.g., Bitcoin. 6 Conclusion The possibility to store non-financial data on cryptocurrency blockchains is both beneficial and threating for its users. Although controlled channels to insert non- financial data at small rates opens up a field of new applications such as digital notary services, rights management, or non-equivocation systems, objectionable or even illegal content has the potential to jeopardize a whole cryptocurrency. Although court rulings do not yet exist, legislative texts from countries such as Germany, the UK, or the USA suggest that illegal content such as child pornography can make the blockchain illegal to possess for all users. As we have shown in this paper, a plethora of fundamentally different meth- ods to store non-financial–potentially objectionable–content on the blockchain exists in Bitcoin. As of now, this can affect at least 112 countries in which pos- sessing content such as child pornography is illegal. This especially endangers the multi-billion dollar markets powering cryptocurrencies such as Bitcoin. To assess this problem’s severity, we comprehensively analyzed the quantity and quality of non-financial blockchain data in Bitcoin today. Our quantitative analysis shows that 1.4 % of the roughly 251 million transactions in Bitcoin’s blockchain carry arbitrary data. We could retrieve over 1600 files, with new con- tent infrequently being added. Despite a majority of arguably harmless content, we also identify different categories of objectionable content. The harmful poten- tial of single instances of objectionable blockchain content is already showcased by findings such as links to illegal pornography or serious privacy violations. Acknowledgements This work has been funded by the German Federal Ministry of Education and Research (BMBF) under funding reference number 16KIS0443. The responsibil- ity for the content of this publication lies with the authors. References 1. German Criminal Code, Section 11 (2013) 2. German Criminal Code, Sections 184b and 184c (2013) 3. Protection of Children Act, Chapter 37, Section 7 (2015) 4. Bitcoin Transaction Fees. https://bitcoinfees.info (2016) Accessed 09/23/2017. 5. General Data Protection Regulation, Section 24 (2016) 6. Aberer, K., Despotovic, Z.: Managing Trust in a Peer-2-Peer Information System. In: ACM CIKM. (2001) pp. 310–317 7. Adya, A., Bolosky, W.J., Castro, M., Cermak, G., Chaiken, R., Douceur, J.R., Howell, J., Lorch, J.R., Theimer, M., Wattenhofer, R.P.: FARSITE: Federated, Available, and Reliable Storage for an Incompletely Trusted Environment. SIGOPS Oper. Syst. Rev. 36(SI) (2002) pp. 1–14 8. Ali, M., Shea, R., Nelson, J., Freedman, M.J.: Blockstack: A New Decentralized Internet. (2017) Accessed 09/23/2017. https://bitcoinfees.info 9. Andresen, G.: Block v2 (Height in Coinbase). https://github.com/bitcoin/ bips/blob/master/bip-0034.mediawiki (2012) Accessed 09/23/2017. 10. Andresen, G.: Pay to Script Hash. https://github.com/bitcoin/bips/blob/ master/bip-0016.mediawiki (2012) Accessed 09/23/2017. 11. Ateniese, G., Magri, B., Venturi, D., Andrade, E.: Redactable Blockchain – or – Rewriting History in Bitcoin and Friends. In: IEEE EuroS&P. (2017) pp. 111–126 12. Bartoletti, M., Pompianu, L.: An analysis of Bitcoin OP RETURN metadata. In: FC Bitcoin Workshop. (2017) 13. Bellinger, J., Hussain, M.: Freedom of Speech: The Great Divide and the Common Ground between the United States and the Rest of the World. Islamic Law and International Human Rights Law: Searching for Common Ground? (2012) pp. 168– 180 14. Blockchain.info: Bitcoin Charts. https://blockchain.info/charts (2011) Ac- cessed 09/23/2017. 15. Camenisch, J., Derler, D., Krenn, S., Pöhls, H.C., Samelin, K., Slamanig, D.: Chameleon-Hashes with Ephemeral Trapdoors. In: PKC ’17. (2017) pp. 152–182 16. Clark, J., Essex, A.: CommitCoin: Carbon Dating Commitments with Bitcoin. In: FC. (2012) pp. 390–398 17. Clarke, I., Sandberg, O., Wiley, B., Hong, T.W.: Freenet: A Distributed Anony- mous Information Storage and Retrieval System. In: Designing Privacy Enhanc- ing Technologies: Workshop on Design Issues in Anonymity and Unobservability. (2001) pp. 46–66 18. Committee to Protect Journalists: Chinese journalist accused of illegally acquiring state secrets. https://cpj.org/x/660d (2015) Accessed 09/23/2017. 19. Damiani, E., di Vimercati, D.C., Paraboschi, S., Samarati, P., Violante, F.: A Reputation-based Approach for Choosing Reliable Resources in Peer-to-peer Net- works. In: ACM CCS. (2002) pp. 207–216 20. Dell Security: Annual Threat Report. (2016) Accessed 09/23/2017. 21. Douglas, D.M.: Doxing: a conceptual analysis. Ethics and Information Technology 18(3) (2016) pp. 199–210 22. Eyal, I., Sirer, E.G.: Majority Is Not Enough: Bitcoin Mining Is Vulnerable. In: FC. (2014) pp. 436–454 23. Fleder, M., Kester, M., Sudeep, P.: Bitcoin Transaction Graph Analysis. (2015) 24. Freedom House: Turkey Freedom of the Press Report. https://freedomhouse. org/report/freedom-press/2016/turkey (2016) Accessed 09/23/2017. 25. Gracie, C.: Hong Kong stages huge National Day democracy protests. http: //www.bbc.com/news/world-asia-china-29430229 (2014) Accessed 09/23/2017. 26. Gupta, M., Judge, P., Ammar, M.: A Reputation System for Peer-to-peer Net- works. In: ACM NOSSDAV. (2003) pp. 144–152 27. Heilman, E., Kendler, A., Zohar, A., Goldberg, S.: Eclipse Attacks on Bitcoin’s Peer-to-Peer Network. In: USENIX Security. (2015) pp. 129–144 28. Herald Union: Copyright infringement by illegal file sharing in Ger- many. http://www.herald-union.com/copyright-infringement-by-illegal- file-sharing-in-germany (2015) Accessed 09/23/2017. 29. HugPuddle: Apertus – Archive data on your favorite blockchains. http:// apertus.io (2013) Accessed 09/23/2017. 30. “Hyena”: Cryptograffiti.info. http://cryptograffiti.info Accessed 09/23/2017. 31. Interpol: INTERPOL cyber research identifies malware threat to virtual curren- cies. https://www.interpol.int/News-and-media/News/2015/N2015-033 (2015) Accessed 09/23/2017. https://github.com/bitcoin/bips/blob/master/bip-0034.mediawiki https://github.com/bitcoin/bips/blob/master/bip-0034.mediawiki https://github.com/bitcoin/bips/blob/master/bip-0016.mediawiki https://github.com/bitcoin/bips/blob/master/bip-0016.mediawiki https://blockchain.info/charts https://cpj.org/x/660d https://freedomhouse.org/report/freedom-press/2016/turkey https://freedomhouse.org/report/freedom-press/2016/turkey http://www.bbc.com/news/world-asia-china-29430229 http://www.bbc.com/news/world-asia-china-29430229 http://www.herald-union.com/copyright-infringement-by-illegal-file-sharing-in-germany http://www.herald-union.com/copyright-infringement-by-illegal-file-sharing-in-germany http://apertus.io http://apertus.io http://cryptograffiti.info https://www.interpol.int/News-and-media/News/2015/N2015-033 32. Irish Office of the Attorney General: Child Trafficking and Pornography Act, Section 2. Irish Statue Book (1998) pp. 44–61 33. Kondor, D., Pósfai, M., Csabai, I., Vattay, G.: Do the Rich Get Richer? An Empirical Analysis of the Bitcoin Transaction Network. PLOS ONE 9(2) (02 2014) pp. 1–10 34. Labs, F.S.: Ransomware: How to Predict, Prevent, Detect & Resond. Threat Response (2016) Accessed 09/23/2017. 35. Le Calvez, A.: Non-standard P2SH scripts. https://medium.com/@alcio/non- standard-p2sh-scripts-508fa6292df5 (2015) Accessed 09/23/2017. 36. Lee, D.: France ends three-strikes internet piracy ban policy. http://www.bbc. com/news/technology-23252515 (2013) Accessed 12/12/2017. 37. Lynch, L.: The Leak Heard Round the World? Cablegate in the Evolving Global Mediascape. In Brevini, B., Hintz, A., McCurdy, P., eds.: Beyond WikiLeaks: Implications for the Future of Communications, Journalism and Society. Palgrave Macmillan UK (2013) pp. 56–77 38. Lyons, K., Blight, G.: Where in the world is the worst place to be a Christian? (2015) Accessed 09/23/2017. 39. Maesa, D.D.F., Marino, A., Ricci, L.: Uncovering the Bitcoin Blockchain: An Analysis of the Full Users Graph. In: IEEE DSAA. (2016) pp. 537–546 40. Matzutt, R., Hohlfeld, O., Henze, M., Rawiel, R., Ziegeldorf, J.H., Wehrle, K.: POSTER: I Don’t Want That Content! On the Risks of Exploiting Bitcoin’s Block- chain as a Content Store. In: ACM CCS. (2016) 41. Matzutt, R., Müllmann, D., Zeissig, E.M., Horst, C., Kasugai, K., Lidynia, S., Wieninger, S., Ziegeldorf, J.H., Gudergan, G., Spiecker gen. Döhmann, I., Wehrle, K., Ziefle, M.: myneData: Towards a Trusted and User-controlled Ecosystem for Sharing Personal Data. In Eibl, M., Gaedke, M., eds.: INFORMATIK, Gesellschaft für Informatik, Bonn (2017) pp. 1073–1084 42. McAfee Labs: Threats Report (December 2016). (2016) Accessed 09/23/2017. 43. McReynolds, E., Lerner, A., Scott, W., Roesner, F., Kohno, T.: Cryptographic currencies from a tech-policy perspective: Policy issues and technical directions. In: Springer LNCS. Volume 8976. (2015) pp. 94–111 44. Meiklejohn, S., Pomarole, M., Jordan, G., Levchenko, K., McCoy, D., Voelker, G.M., Savage, S.: A Fistful of Bitcoins: Characterizing Payments Among Men with No Names. In: IMC. (2013) pp. 127–140 45. Nakamoto, S.: Bitcoin: A Peer-to-Peer Electronic Cash System. (2008) https: //bitcoin.org/bitcoin.pdf. 46. Ober, M., Katzenbeisser, S., Hamacher, K.: Structure and Anonymity of the Bit- coin Transaction Graph. Future Internet 5(2) (2013) pp. 237–250 47. Office of the Law Revision Counsel of the United States House of Representatives: U.S. Code, Title 18, Chapter 110, § 2256 (2017) 48. Okupski, K.: Bitcoin Developer Reference. Technical report (2014) 49. Peerenboom, R.P.: Assessing Human Rights in China: Why the Double Standard. (2005) Accessed 09/23/2017. 50. PoEx Co., Ltd: Proof of Existence. https://proofofexistence.com (2015) Ac- cessed 09/23/2017. 51. Puddu, I., Dmitrienko, A., Capkun, S.: µchain: How to forget without hard forks. IACR Cryptology ePrint Archive 2017/106 (2017) Accessed 09/23/2017. 52. Reid, F., Harrigan, M.: An Analysis of Anonymity in the Bitcoin System. In: Security and Privacy in Social Networks. (2013) pp. 197–223 53. Ron, D., Shamir, A.: Quantitative Analysis of the Full Bitcoin Transaction Graph. In: FC. (2013) pp. 6–24 https://medium.com/@alcio/non-standard-p2sh-scripts-508fa6292df5 https://medium.com/@alcio/non-standard-p2sh-scripts-508fa6292df5 http://www.bbc.com/news/technology-23252515 http://www.bbc.com/news/technology-23252515 https://bitcoin.org/bitcoin.pdf https://bitcoin.org/bitcoin.pdf https://proofofexistence.com 54. Scheller, S.H.: A Picture Is Worth a Thousand Words: The Legal Implications of Revenge Porn. North Carolina Law Review 93(2) (2015) pp. 551–595 55. Selcuk, A.A., Uzun, E., Pariente, M.R.: A Reputation-based Trust Management System for P2P Networks. In: IEEE CCGrid. (2004) pp. 251–258 56. Shirriff, K.: Hidden surprises in the Bitcoin blockchain and how they are stored: Nelson Mandela, Wikileaks, photos, and Python software. http://www. righto.com/2014/02/ascii-bernanke-wikileaks-photographs.html (2014) Ac- cessed 09/23/2017. 57. Sleiman, M.D., Lauf, A.P., Yampolskiy, R.: Bitcoin message: Data insertion on a proof-of-work cryptocurrency system. In: ACM CW. (2015) pp. 332–336 58. Snow, P., Deery, B., Lu, J., Johnston, D., Kirby, P.: Factom: Business Processes Secured by Immutable Audit Trails on the Blockchain. https://www.factom.com/ devs/docs/guide/factom-white-paper-1-0 (2014) Accessed 09/23/2017. 59. Spagnuolo, M., Maggi, F., Zanero, S.: BitIodine: Extracting Intelligence from the Bitcoin Network. In: FC. (2014) pp. 457–468 60. Standing Committee of the National People’s Congress: Law of the People’s Re- public of China on Guarding State Secrets. (1989) Accessed 09/23/2017. 61. Taylor, G.: Concepts of Intention in German Criminal Law. Oxford Journal of Legal Studies 24(1) (2004) pp. 99–127 62. Tomescu, A., Devadas, S.: Catena: Efficient non-equivocation via bitcoin. In: IEEE S&P. (2017) pp. 393–409 63. Tucker, E.: A Look at Federal Cases on Handling Classified In- formation. http://www.military.com/daily-news/2016/01/30/a-look-at- federal-cases-on-handling-classified-information.html (2016) Accessed 09/23/2017. 64. United Nations: Appendix to the Optional protocols to the Convention on the Rights of the Child on the involvement of children in armed conflict and on the sale of children, child prostitution and child pornography (2000) 65. United Nations: Optional protocols to the Convention on the Rights of the Child on the involvement of children in armed conflict and on the sale of children, child prostitution and child pornography. 2171 (2000) pp. 247–254 66. Waldman, M., Rubin, A.D., Cranor, L.: Publius: A Robust, Tamper-Evident, Censorship-Resistant and Source-Anonymous Web Publishing System. In: USENIX Security. (2000) pp. 59–72 67. Walsh, K., Sirer, E.G.: Experience with an Object Reputation System for Peer- to-peer Filesharing. In: NSDI. (2006) 68. Wei, W.: Ancient ’STONED’ Virus Signatures found in Bitcoin Block- chain. https://thehackernews.com/2014/05/microsoft-security-essential- found.html (2014) Accessed 09/23/2017. 69. Wood, G.: Ethereum: A Secure Decentralised Generalised Transaction Ledger. Ethereum Project Yellow Paper (2016) Accessed 09/23/2017. 70. Zeilinger, M.: Digital art as ‘monetised graphics’: Enforcing intellectual property on the blockchain. Philosophy & Technology (2016) 71. Ziegeldorf, J.H., Grossmann, F., Henze, M., Inden, N., Wehrle, K.: CoinParty: Secure Multi-Party Mixing of Bitcoins. In: ACM CODASPY. (2015) pp. 75–86 72. Ziegeldorf, J.H., Matzutt, R., Henze, M., Grossmann, F., Wehrle, K.: Secure and Anonymous Decentralized Bitcoin Mixing. FGCS 80 (3 2018) 448–466 73. Zimmermann, T., Rüth, J., Wirtz, H., Wehrle, K.: Maintaining Integrity and Reputation in Content Offloading. In: IEEE/IFIP WONS. (2016) pp. 1–8 http://www.righto.com/2014/02/ascii-bernanke-wikileaks-photographs.html http://www.righto.com/2014/02/ascii-bernanke-wikileaks-photographs.html https://www.factom.com/devs/docs/guide/factom-white-paper-1-0 https://www.factom.com/devs/docs/guide/factom-white-paper-1-0 http://www.military.com/daily-news/2016/01/30/a-look-at-federal-cases-on-handling-classified-information.html http://www.military.com/daily-news/2016/01/30/a-look-at-federal-cases-on-handling-classified-information.html https://thehackernews.com/2014/05/microsoft-security-essential-found.html https://thehackernews.com/2014/05/microsoft-security-essential-found.html A Quantitative Analysis of the Impact of Arbitrary Blockchain Content on Bitcoin