Friday, February 3, 2017

Security and Claim to Privacy

4
Security and the Claim to Privacy
Keywords: security vs privacy
Louise Amoore
DOI: http://dx.doi.org/10.1111/ips.12044 108-112 First published online: 1 March 2014
ArticleInformation & metricsExplore
PDF
When US President Barack Obama publicly addressed the data mining and analysis activities of the National Security Agency (NSA), he appealed to a familiar sense of the weighing of the countervailing forces of security and privacy. “The people at the NSA don't have an interest in doing anything other than making sure that where we can prevent a terrorist attack, where we can get information ahead of time, we can carry out that critical task,” he stated. “Others may have different ideas,” he suggested, about the balance between “the information we can get” and the “encroachments on privacy” that might be incurred (Obama 2013). In many ways, conventional calculations of security weigh the probability and likelihood of a future threat on the basis of information gathered on a distribution of events in the past. Obama's sense of a trading-off of security and privacy shares this sense of a calculation of the tolerance for the gathering of data on past events in order to prevent threats in the future. In fact, though, the very NSA programs he is addressing precisely confound the weighing of probable threat, and the conventions of security and privacy that adhere to strict probabilistic reasoning. The contemporary mining and analysis of data for security purposes invites novel forms of inferential reasoning such that even the least probable elements can be incorporated and acted upon. I have elsewhere described these elements of possible associations, links, and threats as “data derivatives” (Amoore 2011) that are decoupled from underlying values and do not meaningfully belong to an identifiable subject. The analysis of data derivatives for security poses significant difficulties for the idea of a data subject with a recognizable body of rights to privacy, to liberty, and to justice.

Consider, for example, two years before the controversies surrounding NSA and the PRISM program, when the US Director of National Intelligence, James Clapper, testified before the joint hearing of committees on the virtues of the algorithmic piecing together of fragments of data:

The most valuable national intelligence is the huge collection of databases of routinely collected information that can be searched by computer algorithm. An analyst may know that a potential terrorist attacker is between 23 and 28 years old, has lived in Atlanta Georgia, and has travelled often to Yemen. That analyst would like to be able to very rapidly query the travel records, the customs and border protection service, the investigative records of the State Department … we made important progress after the December 2009 attempted bombing of an aircraft over Detroit, but there remains much more to be done. (US Senate Committee 2011:16)

What is sought in the director of the DNI's vision is not the probable relationship between data on past activities and a future terrorist attack, but more specifically, a potential terrorist, a subject who is not yet fully in view, who may be unnamed and as yet unrecognizable. The security action takes place on the terrain of a potential future on the horizon, a future that can only be glimpsed through the plural relations among data points.

Why does the analysis of fragments of correlated data derivatives matter to our sense of privacy, both as a concept and as a body of rights? Contemporary forms of security do not primarily seek out a complete data set of information on a person, but rather to assemble sufficient correlated data points to make inferences about risk or threat possible. The question of what constitutes personal or private data thus becomes more complex than identifying the accessing of private content. In the public debate on the PRISM program, for example, the idea of “meta-data” mobilizes a distinction between the spoken content of a telephone conversation and the meta-elements of call number, location, time of call, and so on. In fact, though, the complete content of a call event does not need to be captured in order for a security action to take place. The derivative meta-data of associative links between elements represent highly valuable data points for social network analysis techniques that map so-called “patterns of life.” In a sense, the content of the “dots to be connected” matters less to this kind of analysis than the relations between them. The absence of a complete record thus becomes a security virtue because it makes it more difficult to assume that content is innocent—in effect, apparently innocent content can become suspicious at the level of meta-data because it is subsequently associated with other things. For the individual, who is both the subject of privacy rights and the data subject of data protection laws, contemporary security appears indifferent to the person as such, while attentive to the multiple links and associations among plural people and things.

Put simply, contemporary forms of security are less interested in who a suspect might be than in what a future suspect may become; less interested in the one-to-one match of the watch list or alerts index database, and more interested in the signals of real-time predictive analytics. The one-to-one match with an identified individual gives way to what Gilles Deleuze (1992:3) calls the “dividual”—a “torn chain of variables” dividing the subject within herself, recombining with the divided elements of others. The signature of the dividual that is sought by the security software does not “belong” to a subject as such—it does not sign off on past events, it signals possibilities. Thus, for example, an agreement to “depersonalize” data in a database after a period of six months, as in the case of the EU-US agreement on passenger name record (PNR) data, scarcely matters when what is left is a dividuated signature that allows for the writing of new code. So, a subject's specific and named journey on a particular date may no longer be attributable in the database, but it persists as a source for new algorithmic code to act on future subjects.

Among the implications of the data derivative, then, is the capacity for little pieces of past data to become mobile, to detach themselves from identifiable data subjects, and to become active in new security decisions. Some dividuated data elements are disaggregated from the remainder and become drawn into association with other elements derived from other subjects. In this process of disaggregation and reassembly, the person appears less as a singular body than as a plural set of variables which can have an onward life as they connect with the lives of others. For example, the 2009 New York subway suspect, Najibullah Zazi, the 2009 Northwest Airlines Detroit bomber, Umar Farouk Abdulmutallab, and the 2010 Times Square bomber, Faisal Shahzad, could be offered as examples of the failure of systems of data analysis to act upon individuals. Yet, though none of the suspects were preemptively intercepted on the basis of the analysis of their travel or financial data, the fragments of data did reappear as a chain of variables for future algorithmic searches. In the analysis of the US Attorney General, they had ceased to be identifiable subjects and had emerged as “operatives with British or American passports or visas who had visited South Asia and had returned to the US over (redacted) time period” (2010:4). It is the chain of variables of their data fragments that live on in the algorithmic rules of contemporary border security analytics.

If the subject of data-driven security measures is a fractionated subject whose missing elements are a resource to analytics technologies, then can we meaningfully conceive of a right to privacy of the dividual? Such a form of privacy would adhere not to identifiable individuals, but to the associations and links between multiple pieces of multiple people's lives. If what is collected, analyzed, deployed, and actioned in contemporary security is not strictly personal data elements, but a “data derivative”—a specific form of abstraction like a financial derivative, that can have value that is decoupled or only loosely tied to its underlying coordinate—can privacy have meaningful purchase on this slippery derivative? One response has been to seek to restrict the use of personal data and to assert the right to be deleted or forgotten (Mayer-Schönberger 2009). But the data derivative makes associations and links that will always exceed a specified use, will travel, and circulate with new effects and implications. Similarly, the design of new “sovereign information sharing” systems that promise to protect the “integrity of the database” appear to protect individual privacy while allowing their derivatives to circulate (Agrawal, Rakesh, and Ramakrishnan Srikant 2010). By running the analytics within separate databases (airline passenger name record data, national authorities' alerts index data, and so on), and aggregating only the high-risk scores and positive matches in a “master calculation,” the system uses the associations between subjects without revealing the full picture to the data owners. The right to privacy of the individual is thus inadequate to the task of inhibiting the spread of unmoored and dividuated pieces of data that are only partially attributable to a subject.

So, what might it mean to have data protection, or to limit use to a defined purpose in a world of abstracted data derivatives? The gathering of evidence of terrorist activities after the fact is seeping into the preemptive search for indicators of intent capable of anticipating future possible infraction. In this sense, for example, what is significant about the UK's GCHQ generating 197 leads from PRISM in 2012 is not the 197, but how the 197 are arrived at. The 197 represent persons of interest that have come to attention after a process of running large volumes of data through the analytics software. This process is described by the industry as one of “knowledge discovery” in which the algorithm “picks out the strongest relationships” and partitions and segments the data on this basis. Because the knowledge discovery process does not begin with a defined purpose or criteria, it defies the logic of restricting the purpose of data analysis. It is the analytics that govern what the elements of interest should be—this travel associated with this financial transaction, associated with this presence in an open source online community, associated with this presence on Facebook, for example. So, put simply, the vast bulk of the filtering and analysis happens before any named individuals or lists are identified. Can a data subject ever give meaningful consent for their data to be analyzed in bulk, along with terabytes of the data of other people's lives and transactions? Taking seriously the implications of contemporary data analytics, it is not strictly the privacy of the individual that is infringed by programs such as PRISM, but the harm is in the violence done to associational life, to the potentiality of futures that are as yet unknowable (de Goede 2012). In effect, the data do not have meaning until sense is made of them by the relational structure of the analytics. One could protect data very effectively and yet still the inferred meanings of the analytics would limit life chances and close down potential futures.

Perhaps the most significant harm, though, lies in the violence done to politics itself, and to the capacity to make a political claim. Where politics exists because of intractability and irresolvability, because of difficulty as Thomas Keenan (1997) puts it, contemporary data-led security offers resolution through computation, promising to resolve the incalculable and the intractable. The computational and algorithmic turn in security has no tolerance for the emergent or half-seen figure at the edges of perception, no interest in an unfulfilled future potential. “All possible links” are wrought in the correlative relations that are drawn between bodies as they are disaggregated into degrees of risk. There is a challenge for law and human rights when a juridical sensibility of evidence and evaluation meets a set of security measures that demand the projection of future possible events yet to take place. Where the balance of probability may discount unverified fragments, computation lends greater weight to the associated elements, such that once assembled together their singular uncertainty is less easy to see.

So, is privacy as a body of rights adequate to the task of protecting the capacity to make a political claim? If the politics of privacy is to locate space to challenge the analysis of “big data” (Boyd and Crawford 2013), then, one step may be to make a distinction between privacy as a body of rights and privacy as a claim that can be made. The body of rights, as Costas Douzinas (2000, 2012) compellingly reminds us, embodies an imaginary unity of the self, a clearly identifiable subject, while the rights claimant is fractured and split. While the right to privacy itself is readily reconciled within the technology—“anonymizing,” “masking,” or “depersonalizing” the data, for example—the fractured subject falls beneath the register of recognizable rights. What matters politically, then, is that claims to privacy are made, even in the face of the impossibility of meaningful juridical redress. As the layering of software calculations and automated security decisions become ever more difficult to prise open and call to account, the claim to privacy nonetheless illuminates the limit points of what can be made transparent or accountable. Where the search for a definitive body of privacy rights will necessarily fall short of marking a place for all future claims, the act of making a claim is a political demand that can be made by those whose experiences and lives are not registered in an existing body of rights.

© 2014 International Studies Association
References
Agrawal, Rakesh and Ramakrishnan Srikant. (2010) Enabling Sovereign Information Sharing Using Web Services. SIGMOD, Paris, France.
Amoore Louise. (2011) Data Derivatives: On the Emergence of a Security Risk Calculus for Our Times. Theory, Culture and Society 28(6): 24–43.Abstract/FREE Full Text
Boyd Danah, Crawford Kate. (2013) Critical Questions for Big Data. Information, Communication and Society 15(5): 662–679.
Deleuze Gilles. (1992) Postscript on the Societies of Control. October 59: 3–7.CrossRef
Douzinas Costas. (2000) The End of Human Rights: Critical Legal Thought at the Fin-de-Siecle. Oxford: Hart.
Douzinas Costas. (2012) The Paradoxes of Human Rights. Constellations 20 (1): 51–67.
de Goede Marieke. (2012) Speculative Security: The Politics of Pursuing Terrorist Monies. Minneapolis: University of Minnesota Press.
Keenan Thomas. (1997) Fables of Responsibility: Aberrations and Predicaments in Ethics and Politics. Stanford, CA: Stanford University Press.
Mayer-Schönberger Viktor. (2009) Delete: The Virtue of Forgetting in the Digital Age. Princeton, NJ: Princeton University Press.
Obama Barack. (2013) Remarks at White House Press Conference, August 9 2013. Available at http://www.whitehouse.gov/the-press-office/2013/08/09/remarks-president-press-conference. (Accessed November 29, 2013).
US Attorney General. (2010) Review of Passenger Name Records Data. Washington, DC: Office of the Attorney General.
US Senate Committee. (2011) Ten Years After 9/11: Is Intelligence Reform Working?. Washington, DC: US Senate Committee on Homeland Security and Governmental Affairs.

View Abstract

CODE: 44

No comments: