We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

stat.ML

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Machine Learning

Title: Detecting and Quantifying Malicious Activity with Simulation-based Inference

Abstract: We propose the use of probabilistic programming techniques to tackle the malicious user identification problem in a recommendation algorithm. Probabilistic programming provides numerous advantages over other techniques, including but not limited to providing a disentangled representation of how malicious users acted under a structured model, as well as allowing for the quantification of damage caused by malicious users. We show experiments in malicious user identification using a model of regular and malicious users interacting with a simple recommendation algorithm, and provide a novel simulation-based measure for quantifying the effects of a user or group of users on its dynamics.
Comments: Short version, appeared at ICML workshop on Socially Responsible Machine Learning 2021
Subjects: Machine Learning (stat.ML); Cryptography and Security (cs.CR); Machine Learning (cs.LG); Applications (stat.AP)
Cite as: arXiv:2110.02483 [stat.ML]
  (or arXiv:2110.02483v1 [stat.ML] for this version)

Submission history

From: Andrew Gambardella [view email]
[v1] Wed, 6 Oct 2021 03:39:24 GMT (111kb,D)
[v2] Thu, 7 Oct 2021 10:56:01 GMT (111kb,D)

Link back to: arXiv, form interface, contact.