queelius/likelihood.model.series.md

Likelihood model for series systems with masked component cause of failure and other censoring mechanisms

• I’m Alex Towell and I can be reached at [email protected].
• I have two masters degrees from SIUE: Computer Science and Mathematics/Statistics.
• I’m interested encrypted search and homomorphic encryption, oblivious and probabilitistic data structures and algorithms, machine learning and statistics, AI, and programming.
• I’m looking to collaborate on papers (some partially complete). Here are some ideas, but I'm open to other opportunities:
  • Oblivious, privacy-preserving algebraic data types for confidential computation on untrusted systems, with analysis informed by information and probability theory. The data types are algebraic in nature because I have been researching ways to compose them to facilitate building larger oblivious programs from smaller oblivious components, the essence of programming.
  • Probabilistic algorithms and probabilistic algebraic data types primarily concerned with specifying a type of approximation error (normally due to rate distortion) which I tentatively refer to as the Bernoulli Model.
    • Probabilistic data structures that model set-indicator functions, like the Bloom filter, are a well-known special case, but I seek to significantly generalize the results and propagate information about the approximation error through a family of monadic constructions.
    • I have been pursuing derivations of the expected lower-bounds on the space complexity of these approximate Bernoulli types in addition to practical near-optimal data structures that model them.
    • Related to my Computer Science thesis, I have also applied the above results to an approximate Boolean algebra for encrypted search.
  • Reliability engineering and applying statistical inference and learning to predict likely breakdowns (and its causes) of critical systems.
    • It concerns reliability theory and my publication titled "Estimating how confidential encrypted searches are using moving average bootstrap method" concerns reliability engineering.
    • My master's paper "Reliability Estimation in Series Systems: Maximum Likelihood Techniques for Right-Censored and Masked Failure Data" is also related.
  • An information-theoretic model of an optimal adversary (provides a lower-bound on confidientiality in some cases) who, with some probability of success, compromises the confidentiality of an encrypted search system by observing a time series of inputs and outputs.
  • Decentralized "trust machines" (technological solutions to securing trust that does not rely on central authorities), Research on oblivious, privacy-preserving computations is one of the tools in automating trust, but I'm also interested in technologies like Blockchain.