Search

Processing large volumes of data has become increasingly important to businesses and government [218]. One of the popular tools used in processing large sets of data is MapReduce. As new MapReduce applications are developed to be deployed in untrusted environments, such as public clouds, or to process sensitive data or to be deployed across data centres, well understood security measures may be deployed, such as authentication and authorisation or encryption of messages passed between cluster nodes. However, there may be situations where authorised individuals cannot be trusted, such as “rogue” system administrators, or, in the cloud, where the MapReduce cluster nodes have been compromised. Where the input data is sensitive, such as medical data, we require a means to protect this data from exposure. Furthermore, we may need to protect the intermediate data and details of the computation from snoopers to prevent information leakage. To take full advantage of MapReduce cloud computing services, we require a means to process the data securely on such a platform. We designate such a computation, secure computation in the cloud (SCC). SCC should not expose input or output data to any other party, including the cloud service provider. Furthermore, the details of the computation should not allow any other party to deduce its inputs and outputs. Most importantly, we require the computations to be performed in practically reasonable time and space. The ability to perform MapReduce computations on encrypted data would offer a solution to these problems. However, this poses a significant problem in that many encryption schemes transform the original data in such way that meaningful computation on the encrypted data is impossible. Our work aims to provide a practical SSCC system (CryptMR) inspired by CryptDB [272]. Our solution (CryptMR) has detailed several novel cryptographic methods suitable for implementation in a secure computation in the cloud solution using MapReduce. We encrypt integer data using a novel somewhat homomorphic encryption scheme (SHE). This SHE can be made fully homomorphic. We also provide novel order-preserving (OPE) and searchable symmetric encryption (SSE) for the purposes of sorting and searching. Our OPE scheme is, to our knowledge, the first scheme to rely on a computationally hard primitive rather a security model. We have implemented all of these encryption schemes and devised experiments to test their suitability for large-scale distributed computing by integrating them into Hadoop MapReduce (MR) applications. In addition to the work on encryption schemes, we have provided a novel probabilistic method for verifying that mappers and reducers are correctly computing and reporting their results (see chapter 10). This work uses random sampling to minimise the probability that a cheating mapper or reducer can successfully report a false result. We evaluated our “proof-of-concept” implementation in a small scale cloud environment. Our results show that sampling 5 to 10% of intermediate or final data allows us to detect cheating with very strong probability.

The topic of this study is self-tracking. To “self-track” is to quantify steps, calories, heart rate, social interactions, geo-location, posture, and so on. The few monographs, academic papers, and conferences, concerned with this topic often deliver similar philosophical representations of self-tracking based on shared material and socio-cultural assumptions. Generally, these assumptions are wedded to a series of essentialist, absolutist, and deterministic tropes, as well as to neo-humanist and more traditionally humanist forms of discourse. The present study argues that these established discourses play against the predominant tendencies of much contemporary philosophical thought. Contemporary thought is more post humanist than humanist, draws attention to mediation more than media, qualities of indeterminacy more than determinacy, and dynamism more than stasis. The lack of imagination in current discourses of self-tracking, as well as their disengagement with contemporary thought, has resulted in the most pertinent philosophical insight of the phenomena of self-tracking being overlooked: namely the permanence of change. Unique to discourses of self-tracking, this study foregrounds the permanence of change as an already kinetic and transformative aspect of the natural world. As a result of this unique contribution, this study backgrounds the monotony of substance-metaphysical thought as a resistance to the established trends of self-tracking discourses. There are four parts to this process-oriented world view that make up the main features of this research. Firstly, the missing dynamism of self-tracking is rendered actual by its reframing in an event-based ontology, therefore emphasising the emergence, mutation, and inevitable excesses of self-tracking phenomena. Secondly, self-tracking is critically situated in anti-representational terms by focusing on the recent processual-turn in new media theory. Thirdly, via a genealogical method, self-tracking is rearticulated as something dynamic, with a long and plural history of transformative relations. Finally, by emphasising the political, relational, and fictional, a re-imagined speculative alternative to self-tracking is brought to focus through the optic of contemporary design theory.

Introduction & Objectives: There is mounting evidence that expression of the receptor tyrosine kinase, AXL, correlates with poor prognosis in many human cancers. We aimed to evaluate this association in bladder cancer and to examine the in vitro effect of foretinib, an oral inhibitor of VEGF, c-MET and AXL, in bladder cancer cell lines. Material & Methods: Immunohistochemistry was performed on 65 bladder cancer specimens collected from TURBT prior to radical cystectomy (series 1), in addition to a TMA (n=284) comprising bladder cancers of differing stages and grades encompassing both NMIBC and MIBC (series 2). AXL expression was correlated with pathological stage, grade, subsequent need for radical treatment in NMIBC and survival. We used Western blotting and densitometry to evaluate the effects of foretinib on the phosphorylation of signalling molecules downstream of AXL, namely MEK, AKT and MAPK in J82 bladder cancer cells. Results: AXL immunopositivity was associated with invasion in both patient series. In patients undergoing cystectomy, AXL positivity was associated with pre-operatively unsuspected lymph node metastasis. Furthermore, in patients with NMIBC in series 2, AXL expression was associated with an eventual need for cystectomy and with reduced disease-specific survival. Incubation with foretinib (10nM) produced a significant reduction in phosphorylation of MEK and AKT but not MAPK in serum starved J82 cells following stimulation with GAS6, the ligand for AXL. Conclusions: We provide evidence that AXL expression is aligned with the phenotypes of bladder cancers associated with poor outcomes and that AXL signalling can be abrogated with foretinib. Taken together this supports the continued study of AXL inhibitors in bladder cancer.