Rethinking Space in Cybersecurity + ‘Algorithmic Dimensionality’

After initially volunteering to give a ‘lightning’ talk at the CDT in Cyber Security joint conference (Programme) at Royal Holloway next week (3 & 4 May), I was given the opportunity to speak at greater length for 30 minutes. This has provided me the breathing space to consider how I have been conceptualising space in cybersecurity – and is likely to form the basis for the last chapter of my thesis and a subsequent paper I wish to develop out of this (and what I see I’ll be doing post-PhD).

This draws further upon the talk I gave just over a week ago at Transient Topographies at NUI Galway, Ireland. In this, I explored the formation of the software ⇄ malware object, and how this relates to concepts of topography and topology in geography and beyond. In this, I explored how space is thought through in cybersecurity; whether through cartesian representations, cyberspace, or the digital. In my re-engagement with material in software studies and new media, I have intensified the political spheres of my (auto)ethnographic work in a malware analysis lab . Namely, how we come to analyse, detect, and thus curate malware (in public opinion, in visualisations, in speeches and geopolitical manoeuvres) as something that affects security and society. This is not something I claim as anything new, by the way, with Jussi Parikka in Digital Contagions doing this on malware and ‘viral capitalism’, and the multiple works on the relation between objects and security.

Instead, I wish to trace, through my own engagements with malware and security organisations, how space has been thought of. This is in no way a genealogy which would be anything near some contributions (yet) on space and security – but I see this as a start on this path. In particular, how has computer science, mathematics, cybernetics, cyber punk literatures, the domestication of computing, and growing national security anticipatory action conditioned spatial understandings of malware? This has both helpful and unhelpful implications for how we consider collective cybersecurity practises – whether that be by government intervention, paid-for endpoint detection (commonly known as anti-virus) surveillant protection through scanning and monitoring behaviours of malware, attribution, senses of scale, or threat actors – among a variety of others.

Representation of a ‘deep-learning’ neural network. Each layer is connected to each other with different weights, according to each particular application. Some neurons become activated according to certain features.

This working of space in cybersecurity is tied with what I term ‘algorithmic dimensionality‘ in our epoch – where algorithms, and primarily neural networks, produce dimensional relations. What I mean by dimensions is the different layers, that come together to produce certain dimensions of what to follow at each consecutive layer, generating relationships that are non-linear; that can be used for malware detection, facial recognition, and a variety of other potential security applications. These dimensions exist beyond humanly comprehension; even if we can individually split neuron layers and observe what may be happening, this does not explain how the layers interact adequately. Hence, this is a question that extends beyond, and through, an ethics of the algorithm – see Louise Amoore‘s forthcoming work, which I’m sure will attend to many of these questions – to something that is more-than-human.

We cannot simply see ethics as adapting bias. As anyone who has written neural networks (including myself, for a bit of ‘fun’), weights are required to make it work. Algorithms require bias. Therefore reducing bias is an incomplete answer to ethics. We need to consider how dimensionality, which geographers can engage with, is the place (even cyberspatial) in which decisions are made. Therefore, auditing algorithms may not be possible without the environments in which dimensionality becomes known, and becomes part of the generation of connection and relationality. Simply feeding a black box and observing its outputs does not work in multi-dimensional systems. Without developing this understanding, I believe we are very much lacking. In particular – I see this as a rendition of cyberspace – that has been much vented as something that should be avoided in social science. However dimensionality shows where real, political, striations are formed that affect how people of colour, gender, and sexual orientation become operationalised within the neural network. This has dimensional affects that produce concrete issues; whether by credit ratings, adverts shown, among other variables that are hard to grasp or prove discrimination.

Going back to my talk at Royal Holloway (which may seem far from the neural network), I will attempt to

 enrol this within the conference theme of the ‘smart city’, and how certain imaginaries (drawing heavily from Gillian Rose’s recent thoughts on her blog) are generative of certain conditions of security. By this, how do the futuristic, clean, bright images of the city obscure and dent alternative ways of living and living with difference? The imaginings of space and place, mixed with algorithmic dimensionality, produce affects that must be thought of in any future imagining of the city. This draws not only from my insight from my PhD research on malware ecologies, in which I attempt to open-up what cybersecurities are and should include (and part of an article I am currently putting together), but also include feminist and queer perspectives to question what the technologically-mediated city will ex/in/clude.

I think space has been an undervalued concept in cybersecurity. Space and geography has been reduced to something of the past (due to imaginaries of the battlefield disappearing), and something that is not applicable in an ‘all-connected’ cyberspace. Instead, I wish to challenge this and bring critical analysis to cyberspace to explore the geographies which are performed and the resultant securities and inequalities that come from this. This allows for a maturity in space and cybersecurity – that appreciates that space is an intrinsic component of interactions at all levels of thinking. We cannot abandon geography, when it is ever more important in securities of everyday urban areas, in malware analysis, geopolitics, and even in the multi-dimensionality of the neural network. Hence space is an important, and fundamental, thing to engage with in cybersecurity which does not reduce it to the distance between two geometrically distant places.


Leave a Reply