Decentralized Reputation System for Transaction Networks Dept. of CIS - Senior Design 2014-2015∗ Matthew Buechler Manosai Eerabathini Christopher Hockenbrocht [email protected] [email protected] [email protected] Univ. of Pennsylvania Univ. of Pennsylvania Univ. of Pennsylvania Philadelphia, PA Philadelphia, PA Philadelphia, PA Defu Wan [email protected] Univ. of Pennsylvania Philadelphia, PA ABSTRACT of a person's trustworthiness by others. In any interaction Peer-to-peer transaction networks such as Bitcoin have gained between two or more parties, reputation drives the ability traction in the past few years as they are touted as an anony- for each party to trust one another and facilitate a success- mous, safe, and secure way to make electronic payments with ful experience. By nature, reputation systems have been low transaction fees. Our goal is to improve these technolo- centralized with application developers defining their own gies by developing a reputation system that calculates reputa- standard of reputation in the context of their own use case. tion based on underlying network structure and allows users For example, a site like Quora would come up with a propri- to look up and record histories of transaction outcomes. We etary algorithm to determine the reputation score of every believe that this will facilitate a more honest market envi- Quora user but this reputation would not be the same as ronment. Beyond standard market transactions, developers a user's reputation on a site like Yelp. The landscape of will also be able to quickly build reputation components into reputation systems is fragmented with every company own- their applications by utilizing our work. Our main contri- ing a \secret sauce" that helps it make money by locking in butions comes in two parts: a reputation algorithm called users. Though this comes with strategic advantages for the net flow convergence and a decentralized system that records companies, it requires that a consumer have multiple repu- transaction outcomes. tations stored across different services on the web. Reputa- tion is inherently tied to identity and we believe there is a need to detach reputation from one specific company or site. 1. INTRODUCTION Rather, consumers should be able to determine reputation There is a growing demand for more aspects of the mod- in a distributed manner. We hope to build a decentralized ern Internet to be decentralized. Though Internet applica- reputation system that analyzes both the underlying trans- tions are built on top of decentralized protocols like TCP/IP action network structure and builds a history of transaction and HTTP, a large portion of the application stack remains outcomes to eliminate the need for 3rd parties to define what centralized. Much of the desire for more decentralized com- constitutes good and bad reputation. puting systems comes from concerns regarding mass surveil- lance over the web and limited access to information due to 1.2 Advantages censorship. Only 25 years old, the Internet is amidst an- This kind of reputation system will enable distinct advan- other radical shift in which open-source, decentralized tech- tages such as helping e-commerce businesses and Internet nologies are gaining traction in many aspects of application marketplaces develop faster by using a preexisting reputa- development. tion system. Software developers will no longer need to en- code their own reputation systems from scratch but rather 1.1 Motivation can leverage ours to serve the same purpose. Additionally, Historically, there are many online industries where we can reputation would be a global measure so that a user's rep- observe decentralized computing systems uprooting existing utation is not limited to one particular app or service. A businesses. With the rise of Bitcoin technology, industries decentralized reputation system will protect the consumer like online commerce have also seen decentralized alterna- from being exploited by exterior forces such as governments tives emerge such as Silk Road and OpenBazaar. The suc- or corporations. In general, using decentralized networks cess of these projects has resulted from the groundbreaking ensure the most ideal form of democracy, where the voice ability to transfer money online without requiring a third of the individual can be guaranteed through cryptographic party to verify transactions. Bitcoin technology enables security. Decentralizing the notion of reputation has not many such applications to support a variety of use cases been done in the current technological landscape and there in a decentralized manner. is great value in providing this ability to app developers. Payment, e-commerce, identity, and file-sharing are just a few areas that have been touched upon with decentralized 2. RELATED WORK technologies. Reputation can be defined as the perception Our work touches on a wide variety of research projects ∗Advisor: Jonathan M. Smith([email protected]). and academic papers. We studied other reputation algo- rithms to understand their strengths and weaknesses and useful to score nodes in a network so that we can use them use them as benchmarks. We built on top of decentralized as a baseline comparison to our algorithm. application platforms. We integrated ideas from other areas of decentralized applications. One of them is identity and 2.3 Bitcoin the OpenID project, which is an open standard and decen- A crucial technology for allowing peers in a network to ex- tralized protocol that allows users to be authenticated by change value in a trustless, electronic manner is Bitcoin[8]. certain co-operating sites using a third party service[4]. In Before this innovation, peers would have to use a non-electronic other words, it is a way for users to hold credentials with- form of payment, cash, without a decentralized ledger that out a third party having complete control over their use. handled the accounting computation, or a centralized form Reputation is inherently linked to identity as it cannot ex- of payment, credit cards, that would introduce a vital de- ist without identity. When it comes to sharing data with pendence on a third party and a point of failure in the decen- peers in a network, Persona, a decentralized social network tralized system[8]. We hope to provide two ways to interact with user-defined privacy, is our guide[1]. Namecoin is the with Bitcoin–first providing an easy way for apps to inter- decentralized public registry for identities that utilizes this act and make transactions with other devices and second, concept and we follow its implementation closely. This al- through graph analysis, provide naive measures of reputa- lows us to store already researched cryptographic tools, such tion. Giving apps an abstracted way to interact with Bitcoin as elliptic curve public key signatures which serve as identi- will alleviate the problems of interacting with a complicated fication tools. Attribute-based encryption is a relatively re- infrastructure. cent area of study in cryptography, which might allow such The Bitcoin transaction market is analyzed by the repu- a sharing of information based on attribute or access levels tation algorithm to ascertain the issuance and verification and will allows users in a network to allow for fine-grained of reputation in the network. The peer to peer computing access control of their data[5]. Of course, being a somewhat network that will compute this reputation will be based on new area of research, this might not be possible. protocols that are extensions of the Bitcoin protocol tech- nology. 2.1 EigenTrust EigenTrust is a reputation management algorithm that 2.4 Counterparty calculates reputation as a global trust value in peer to peer Counterparty is a protocol extension built on top of the networks.[6]. The aim was for the algorithm to be applied Bitcoin protocol. It allows for issuing assets and tokens as in file upload networks like Gnutella so malicious users who well for building decentralized exchanges and trustless bet- were uploading inauthentic files could be weeded out based ting[7]. While the processes are routed through the Bitcoin on their score. protocol, Counterparty uses an external native token issued The algorithm itself relies on building transitive trust by through an initial proof-of-burn IPO that acts as a fuel for examining a peer's history of uploads. A node A trusts some any action within the protocol. Using a native token other peer node B so therefore node A will trust any nodes trusted than Bitcoin allows the protocol more flexibility in its evolu- by B. The algorithm works by having node A calculate a tion while not detracting from the existing security benefits local trust value for all of its neighbors and then normalizing of the Bitcoin network in its current form. this value to prevent bad neighbors from colluding with their 2.5 Ethereum cohorts. All local trust values are then aggregated and a trust vector for the whole network is created. Ethereum is an entirely new blockchain with different proof This algorithm is very relevant to the premise of our net- of work and mining security. Its main innovation is that it work convergence algorithm. Flow of money and file uploads includes a Turing complete programming language for devel- can be abstracted in a similar way. However, EigenTrust fo- oping applications within the network[3]. These applications cuses on a reputation system that relies on user feedback, take the form of smart contracts, which are executable pro- but we believe that monetary transactions can themselves grams, and decentralized autonomous organizations, which provide a useful measure of reputation without any user in- are sets of executable programs that work in tandem to pro- put. vide value in the economy. Ethereum includes a native to- ken, ether, distributed in an initial IPO and then mined, that 2.2 Centrality as a Proxy for Reputation Mea- acts as "gas" necessary for executing smart contracts[3].