[LOGOS] LOGOS Meetup this Wed, talk by Ben Treves on Cross-Platform Profiling

[Emiliano De Cristofaro]

Hi Everyone,

Please find the details of our next meetup below.

Cheers,
Emiliano

SPEAKER
Ben Treves, PhD @ UCR CSE

WHEN
Wed Oct 30th, 2pm

WHERE
Bourns A171 (preferred)
Zoom (link
<https://ucr.zoom.us/j/93797597205?pwd=k14o0hy6qk0HnLB99yYCOXCgSlMbrl.1>)

TITLE
Cross-Platform Disambiguation and Profiling of Online Users

ABSTRACT
What can we discover about online users by analyzing the content they post
across platforms? Traditional works primarily focus on eliciting and
analyzing profiles based on single platforms. To this end, we develop
systematic methods for disambiguating and profiling users across platforms,
paying special attention to their URL posting behavior. We first develop
RURLMAN, a modular ensemble of methods for leveraging user-posted URLs to
connect forum users with their cross-platform profiles. Our approach has
two key features: (a) we focus on user-posted URLs as the key source of
information, and (b) we utilize a stacked ensemble integrating multiple
methods, including NLP and LLM capabilities. We extensively evaluate
RURLMAN and deploy it in a case study where we connect users to external
profiles, even linking some users to multiple off-site accounts. Next we
present VIKI, a systematic methodology for extracting and integrating
users’ displayed personas across social networks. The key novelty of our
method lies in our cross-platform approach, as we mine different types of
platforms and extract multiple types of information for each user,
including displayed personality traits, interests, and offensiveness. From
an algorithmic perspective, we evaluate, combine, and introduce methods to
analyze and visualize data, including LLMs. We evaluate VIKI on a dataset
combining three platforms: Github, LinkedIn, and X (Twitter). We see our
work as a key building block towards systematic and nuanced profiling of
users by leveraging their digital footprint across platforms.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ucr.edu/pipermail/logos/attachments/20241028/e02d451b/attachment.html>