Domain-Based Latent Personal Analysis (LPA) and Its Uses

LPA is an easy-to-use and fast domain-based spectral signature that can be used in a variety of domains and is tailored for big-data computations. It creates a strong and unique signature of both over-used and missing items that identify a user or a component in a domain. It can be used in a variety of domains, from textual to computational.

Paper: Mokryn, O., & Ben-Shoshan, H. (2020). “Domain-based Latent Personal Analysis and its use for impersonation detection in social media.” Accepted to User Modeling and User-Adapted Interaction (UMUAI), May 2021. arXiv:2004.02346. (PDF)

Zipf’s law defines an inverse proportion between a word’s ranking in a given corpus and its frequency in it, roughly dividing the vocabulary into frequent (popular) words and infrequent ones. Here, we stipulate that within a domain an author’s signature can be derived from, in loose terms, the author’s missing popular words and frequently used infrequent words. We devise a method, termed Latent Personal Analysis (LPA), for finding such domain-based personal signatures. LPA determines what words most contributed to the distance between a user’s vocabulary and the domain’s.

We identify the most suitable distance metric for the method among several and construct a personal signature for authors. We validate the correctness and power of the signatures in identifying authors and utilize LPA to identify two types of impersonation in social media: (1) authors with sockpuppet (multiple) accounts; (2) front-user accounts, operated by several authors. We validate the algorithms and employ them over a large-scale dataset obtained from a social media site with over 4000 accounts, and corroborate the results employing temporal rate analysis. LPA can be used to devise personal signatures in a wide range of scientific domains in which the constituents have a long-tail distribution of elements.

Impersonation on social media: sockpuppet accounts

LPA is fast and easy to implement at large scale. We deployed it over 4000 IMDb reviewer accounts to find sockpuppet accounts activated by a single author, performing a 4000x4000 similarity measure between all accounts’ LPA signatures. The full list of suspected sockpuppets is available.

Example 1: User 59775972 (joshuadrake-39480) and User 62431316 (joshuadrake-91275) have 502 common terms in their LPA signatures, and the distance between their signatures is 0.013.

Example 2: User 24051675 (jpachar82) and User 53564354 (jasonpachar) have 584 common terms in their LPA signatures, and the distance between their signatures is 0.297.