Cyberspace Governance:

Transcription

1 Cyberspace Governance: security, surveillance and privacy: metadata Virgilio Almeida Harvard University Cambridge, October 2016

2 References YA De Montjoye, CA Hidalgo, M Verleysen, VD Blondel, Unique in the crowd: The privacy bounds of human mobility Nature ScienMfic Reports, March YA De Montjoye, L Radaelli, VK Singh, Unique in the shopping mall: On the re- idenmfiability of credit card metadata, Science Vol. 347, Issue 6221, Hal Abelson et al., Keys under Doormats: MandaMng insecurity by requiring government access to all data and communicamons, Journal of Cybersecurity, Vol. 1(1) (2015). Virgilio Almeida, Doneda, Danilo and Marilia Monteiro, Governance Challenges for the Internet of Things, IEEE Internet CompuMng, vol. 19, no. 4, 2015

3 Course Outline The central goal of the course is to present students a combinamon of knowledge from mulmple disciplines in order to allow them to carry out evidence- based analysis of problems and issues related to cyberspace governance. Weeks : Cyberspace and Internet Governance Concepts a. The development of Internet, the main principles governing Internet b. The Regime Complex for Managing Global Cyber AcMviMes c. The Convergence of Social and Technological Networks d. Internet governance concepts and models e. ICANN and management of domain names f. IANA and management of IP addresses, the case of IPv6

4 Examples of the data-oriented approach Science and Nature armcles Privacy in the Crowd: the privacy bounds of human mobility Privacy in the Shopping Mall: on the reindenmfiability of credit card metadata Propose Changes to Policies Policy Issue: PRIVACY Collect Data: Mobility & Financial Analyze QuanHtaHve Results Create Empirical Data Models for Re- idenhficahon

5 Data can shed light on some key problems in cyberspace governance QuanMtaMve Research to Analyze Public and Private Policies for Cyberspace Development of Regulatory Models Interdisciplinary Studies Empirical Models AnalyMcal Models DATA

6 YA de Montjoye, L Radaelli, VK Singh, AS Pentland Science 347 (6221), UNIQUE IN THE SHOPPING MALL

7 Characteristics of the work It uses 3 months of credit card records for 1.1 million people It shows that four spamotemporal points are enough to uniquely re- idenmfy 90% of individuals. It shows that knowing the price of a transacmon increases the risk of reidenmficamon by 22%, on average. In science, it is essenmal for the data to be available and shareable. Sharing data allows scienmsts to build on previous work, replicate re- sults, or propose alternamve hypotheses and models. These metadata are generated by our use of technology and, hence, may reveal a lot about an individual. Making these data sets broadly available, therefore, requires solid quanmtamve guarantees on the risk of reidenmficamon, à privacy risks

8 Characteristics of the work: metrics Unicity quanmfies the intrinsic re- idenmficamon risk of a data set. Unicity quanmfies how much outside informamon one would need, on average, to re- idenmfy a specific and known user in a simply anonymized data set. The higher a data set s unicity is, the more reidenmfiable it is. Outside informamon could be a tweet that posimons a user at an approximate Mme for a mobility data set or a publicly available movie review for the Nehlix data set. A recent survey shows that financial and credit card data sets are considered the most sensimve personal data worldwide

9 Data Set To provide a quanmtamve assessment of the likelihood of idenmficamon from financial data, we used a data set D of 3 months of credit card transacmons for 1.1 million users in 10,000 shops The data set was simply anonymized, which means that it did not contain any names, account numbers, or obvious idenmfiers.

10 Results The risk of re- idenmficamon of D is quanmfied by unicity ε. Unicity is the risk of re- idenmficamon knowing p pieces of outside informamon about a user. It evaluates ε(p) of D as the percentage of its users who are re- idenmfied with p randomly selected points from their financial trace. Example: Suppose you are searching for Sco1 in a simply anonymized credit card data set. You know two points about Sco1: he went to the bakery on 23 September and to the restaurant on 24 September. Searching through the data set reveals that there is one and only one person in the enare data set who went to these two places on these two days. S(I p ) is thus equal to 1, Sco1 is re- idenafied, and you now know all of his other transacaons, such as the fact that he went shopping for shoes and groceries on 23 September, and how much he spent. Financial traces contain one addimonal column that can be used to reidenmfy an individual: the price of a transacmon. A piece of outside informamon, a spamotemporal tuple can become a triple: space, Mme, and the approximate price of the transacmon.

11 Results: metrics and plots

12 Results: metrics and plots

13 Unique in the Crowd

14 Science versus Art: in the context of cyberspace governance Science Principles Fundamental laws ExplanaMon Discovery Analysis DissecMon Art PracMce Skilled performance AcMon InvenMon Synthesis construcmon To understand informamon processes, computer scienmsts must observe phenomena, formulate explanamons, and test them. This is the scienmfic method Art refers to useful practices of a field. (*)Walter Tichy, Should Computer Scientists Experiment More?

15 Comments on the methods SCIENCE OBJECTS DOMINATING METHOD Logic & Mathematics Simple Abstract objects: propositions, numbers, Reductionism (analysis) Deduction Internet cyberspace Natural Sciences Social Sciences Humanities Natural objects: physical bodies, fields and interactions, living organisms Social objects: human individuals, groups, society, Cultural objects : human ideas, actions and relationships, language, artifacts Hypothetico-deductive method Hypothetico-deductive method + Hermeneutics Hermeneutics, Theory & methodology of interpretation (*)Gordana DODIG- CRNLOVIC ScienMfic Methods in Computer Science

16 Comments on the methods An example of a research quesmon: What privacy means in the digital age where everything is recorded and we leave permanent traces of our idenaty everywhere, whether we like it or not. So what is the relaaon between metadata and privacy? (*)Gordana DODIG- CRNLOVIC ScienMfic Methods in Computer Science

17 The Scientific Method Research question? ExisMng Knowledge 1 Theory SelecMon 6 Hypothesis must be adjusted Hypothesis 2 Hypothesis must be adjusted Old or New Theory confirmed 5 PredicMon 3 Tests & ObservaMon 4

18 Examples of the data-oriented approach Privacy in the Crowd: the privacy bounds of human mobility Privacy in the Shopping Mall: on the reindenmfiability of credit card metada Policy Issue: PRIVACY Propose Changes to Policies Collect Data: Mobility & Financial Analyze QuanHtaHve Results Create Empirical Data Models for Re- idenhficahon

19 Next Class 1. One or two paragraphs describing areas of interest (i.e., potenmal problems or topics) for the course project. I will review the proposals and suggest references. 2. Chapters 5 of DeNardis Book 3. Anne Edmundson, Roya ENsafi, Nick Feamster, Jennifer Rexford,Characterizing and Avoiding RouMng Detours Through Surveillance States, May Hal Abelson et al., Keys under Doormats: MandaMng insecurity by requiring government access to all data and communicamons, Journal of Cybersecurity, Vol. 1(1) (2015).

20 THANKS!