3 years Postdoc position at TCD (Dublin) on Security & Privacy of unused parameters of ML models

72 views
Skip to first unread message

bena...@gmail.com

unread,
Oct 29, 2025, 12:34:55 PM (24 hours ago) Oct 29
to Machine Learning News

Dear colleagues and students,

I am recruiting a Postdoctoral researcher with a background in game theoretic modelling, learning theory and both the theoretical and practical aspects of deep learning. The successful applicant will collaborate within a skilled and motivated team of scientists (from Trinity College Dublin, Queen's University Belfast and University California Riverside) dedicated to understand the security and privacy implications of the unused parameters present in machine learning models. The proposed research explores the problem both from the theoretical and empirical perspectives.

Description: Machine Learning applications increasingly rely on Deep Neural Networks (DeepNNs) to achieve state-of-the-art performance across a wide range of real-world tasks, including Large Language Models. DeepNNs are typically highly over-parameterised, but this characteristic has been shown to facilitate both optimisation and generalisation. Nevertheless, such over-parameterisation also introduces many parameters that are not essential to the final learned model and whose specific values have little impact on its behaviour (don’t-care states). In both software and hardware systems, undefined behaviour and don’t-care states have been shown to be potential sources of vulnerabilities. The postdoc will conduct research that contributes to the following streams of work: 1) Investigate the possible strategies that colluding parties could use to maliciously exploit the unused parameters (learning capacity) of a machine learning model using techniques from information and learning theory; 2) Model the attacks within a game-theoretic framework under complete or incomplete knowledge (Bayesian games), which will enable to find optimal strategies for the players (Attacker, Nature and Defender); 3) Validate these theoretical models and strategies on machine learning models and real applications.

Applications from candidates with expertise in imprecise Statistics or imprecise Machine Learning are welcome.

  • For informal inquiries please contact alessio....@tcd.ie
  • Salary is up to ~€71,000 per annum per 3 years.
  • Start: January 2026.
  • For additional details (including qualifications, knowledge/experience  and how to apply) see attached document.
 Best regards,
Alessio Benavoli
Job Description - Security and Privacy in Machine Learning.docx
Reply all
Reply to author
Forward
0 new messages