Pay as you Browse: Microcomputations as Micropayments in Web-based Services Ghassan O. Karame Aurélien Francillon Srdjan Capkunˇ Dept. of Computer Science Dept. of Computer Science Dept. of Computer Science ETH Zurich, Switzerland ETH Zurich, Switzerland ETH Zurich, Switzerland [email protected] [email protected] [email protected] ABSTRACT increase their revenues. This shift is, however, expected to Currently, several online businesses deem that advertising alienate a considerable number of online users. revenues alone are not sufficient to generate profits and are While users might be willing to pay for low-cost special- therefore set to charge for online content. In this paper, we ized online products such as music and movies, they are not explore a complement to the current advertisement model; keen on accepting subscription charges to read online news, more specifically, we propose a micropayment model for non- to sign in Facebook, etc. In fact, studies have shown that specialized commodity web-services based on microcompu- only a small fraction of users—almost three percent—are tations. In our model, a user that wishes to access online willing to pay to read online news [1, 3]. Users are also not content offered by a website does not need to register or pay willing to set up and frequently recharge accounts for each to access the website; instead, he will accept to run micro- online commodity service that they use. These issues make computations on behalf of the website in exchange for access many commodity websites reluctant to charge for content to the content. These microcomputations can, for example, and/or registration. The challenge for most media and on- support ongoing computing projects that have clear social line businesses lies, therefore, in extracting revenues from benefits (e.g., projects relating to HIV, dengue, cancer, etc.) their online content without alienating existing users. or can contribute towards commercial computing projects. In this paper, we consider this problem and we propose a We argue that this micropayment model is economically and new framework that enables websites to“charge”for content, technically viable and that it can be integrated in existing thus increasing their revenues, without requiring subscrip- distributed computing frameworks (e.g., the BOINC plat- tion charges from their users. Our scheme somehow departs form). We implement a preliminary prototype of a system from current micropayment methods and offers online busi- based on our model through which we evaluate its perfor- nesses an indirect form of remuneration—similar to the cur- mance and usability. Finally, we analyze the security and rent advertisement model. In our scheme (Figure 1), a user privacy of our proposal and we show that it ensures payment wishing to access online content offered by a website does for the content while preserving the privacy of users. not need to register or pay to access the website; instead, he will accept to run some computations on behalf of the website in exchange for access to the content. After verify- 1. INTRODUCTION ing the integrity of the results reported by the user, results In the last couple of years, the drop in Internet advertising of the computations are gathered by the service provider revenues [1,20] has generated a discussion on possible alter- (or by a broker) and sent to a distributed computing part- natives that will increase the revenues of online businesses ner in exchange for a payment. The computations carried and shareholders. Some researchers [20] argue that the loss out by the user could correspond to those used in the mul- in Internet revenues was not solely caused by the general titude of available distributed computing platforms (such economic recession but also finds roots in the online adver- as SETI@home [4], distributed.net [5], etc.); alternatively, tisement model itself. Studies have reported that users do these computations could also be performed on behalf of not trust online advertisements [2]; some users further use governmental agencies, research labs and private industries. tools to block them. Note that, similarly to the existing advertisement model, a Given this, several websites are set to charge for online third party could mediate the exchange between online ser- content. For instance, News Corporation has declared that vices and their computing partners. However, unlike the it will start charging for news content by 2011 [1]. Several targeted advertisement model where the (privacy-invasive) other online businesses are likely to follow the same move to user profiling increases revenues, our framework does not require the content providers to intrude on the privacy of users. Note that our scheme can be used as a complemen- tary model to the existing advertisement model. In this respect, our proposed scheme shares similarities Permission to make digital or hard copies of all or part of this work for with “parasitic computing” [19, 33], where covert computa- personal or classroom use is granted without fee provided that copies are tions are executed on users’ machines without their con- not made or distributed for profit or commercial advantage and that copies sent and knowledge. Our scheme, on the other hand, ex- bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific tends the notion of “parasitic computing” to offer users a permission and/or a fee. transparent—yet undisguised—micropayment method. The Copyright 200X ACM X-XXXXX-XX-X/XX/XX ...$10.00. 17 10 16 10 15 10 14 FLOPS Acquired 10 Online Newspapers Emails+Facebook+YouTube+Online Newspapers 13 10 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Percentage of Online Users that Adopt the Micro−Computations Scheme Figure 2: Total Number of FLOPS acquired in our scheme versus the number of Online Newspa- per Viewers. For comparison purposes, the GIMPS project [6] aggregates around 44 TFLOPS (44 · 1012 FLOPS). accessing a specific online service. Finally, we performed a Figure 1: Example of using microcomputations as a preliminary user study to assess the usability of our model. payment scheme in online newspapers. Upon access- The remainder of this paper is organized as follows. In ing the news website, the client’s browser performs Section 2, we describe the main intuitions behind our pro- microcomputations. The results are then sent back posed scheme. In Section 3, we present our framework and to interested third party in exchange of a payment. we analyze its security and privacy implications. We imple- ment our scheme and we assess its performance in Section 4. In Section 5, we discuss additional insights to our proposals. main benefits of our scheme are as follows: (i) our scheme In Section 6, we overview related work in the area and we is likely to boost the revenues of online services; in fact, conclude the paper in Section 7. such a barter is likely to be more accepted by users—when compared to subscription charges—by exploiting their will- ingness to aid ongoing projects that have clear benefits (e.g., projects relating to HIV, cancer, clean energy, etc.). In 2. MOTIVATION addition, our scheme enables users to state which compu- To better illustrate the benefits of our microcomputations tations/projects they prefer to support. (ii) Due to the scheme, we consider an example where online newspapers increasing reliance on distributed computing platforms to would require their clients to perform computations for as solve computationally challenging projects, our scheme will long they access their content (i.e., their portal). enable existing platforms to harness the idle computing power For the sake of this example, we assume that there are of users for as long as these users are using online services 60 million unique clients accessing online news per month, (e.g., while users are reading online news). and that each client spends on average 50 minutes per month (iii) This entire process can be made transparent to the browsing the news sites [7]. Although the pricing for the cost users as the computations can be carried out within their of computations is well understood (ranging between 0.1 and browsers. (iv) Finally, the advantage of malicious users in 0.3 $ per hour [8,9]), we rely in this example on a worst case mounting attacks against our scheme is negligible since the analysis and we estimate the market cost of computations by targeted service corresponds to a low-cost commodity con- the cost of electricity consumption of the machines involved tent. in the computations. On average, a computer is estimated Note that a thorough analysis of the economic viability of to use 100 watts per hour [10] and the cost of a kWh is our scheme is out of the scope of the paper; instead we show estimated to be 0.1 $ [11]. in this work the feasibility of using microcomputations as a This suggests that the cost of computations can be lower- micropayment method for online content. bounded by 0.01 $ per hour. In addition to proposing the use of microcomputations as In that case, online newspapers will be able to generate a micropayment scheme, we make the following contribu- 1.5 million $ of revenues per month based on our proposed tions. We show that our proposed micropayment scheme model; when compared to online advertisements, this rev- naturally supports the anonymity and the privacy of users. enue is equivalent to each of the 60 million users clicking on Nevertheless, we show that the authenticity and the correct- almost 1 ad per month in the Google AdSense advertisement ness of the results submitted by the user can be verified— framework [15].