Piotr (Peter) Mardziel

About

I work on transparency and accountability in machine learning with applications to security, privacy, and fairness. I work on making computers explain themselves. I want to make sure computers do not learn sad things from the sad world or remember things about us we would rather they not. This makes the world not as sad and computers not as bad. Woof woof. I also dabble in probabilistic programming and quantitative information flow. My interests include:

S&P Topics:

• Anonymity
• Language-based security
• Machine learning and AI security
• Privacy technologies and mechanisms
• Secure information flow

CSF Topics:

• accountability
• anonymity and privacy
• formal methods and verification
• information flow control
• language-based security
• machine learning
• software security

CMU Topics:

• Security and Privacy: Foundations and Formal Methods
• Application Domains: Machine learning
• Data Science: Machine Learning
• Data Science: Deep Learning

Small-ish questions

• How much pi does gpt know?
• Drift Metrics: How to Select the Right Metric to Analyze Drift
• Drift in Machine Learning: Why it's hard and what to do about it

Courses

• Trustworthy Machine Learning (Stanford CS 329T Spring 2021)
• Security and Fairness of Deep Learning (CMU 18-739 Spring 2020)
• Foundations of Privacy (CMU 18-734 Fall 2018)

Papers

• (html)Gender Bias in Neural Natural Language Processing.
  Logic, Language, and Security: Essays Dedicated to Andre Scedrov on the Occasion of His 65th Birthday. 2020.
  Kaiji Lu, Piotr Mardziel, Fangjing Wu, Preetam Amancharla, Anupam Datta
• (html)Probabilistic Abstract Interpretation: Sound Inference and Application to Privacy.
  Foundations of Probabilistic Programming, Cambridge University Press. 2020.
  José Manuel Calderón Trilla, Michael Hicks, Stephen Magill, Piotr Mardziel, Ian Sweet
• Smoothed Geometry for Robust Attribution.
  Neural Information Processing Systems (NeurIPS). 2020.
  Zifan Wang, Haofan Wang, Shakul Ramkumar, Matthew Fredrikson, Piotr Mardziel, Anupam Datta
• Influence Paths for Characterizing Subject-Verb Number Agreement in LSTM Language Models.
  Annual Conference of the Association for Computational Linguistics (ACL). 2020.
  Kaiji Lu, Piotr Mardziel, Klas Leino, Matthew Fredrikson, Anupam Datta
• Interpreting Interpretations: Organizing Attribution Methods by Criteria.
  CVPR Workshop on Fair, Data Efficient and Trusted Computer Vision. 2020.
  Zifan Wang, Piotr Mardziel, Anupam Datta, Matthew Fredrikson
• Score-CAM: Score-Weighted Visual Explanations for Convolutional Neural Networks.
  CVPR Workshop on Fair, Data Efficient and Trusted Computer Vision. 2020.
  Haofan Wang, Zifan Wang, Mengnan Du, Fan Yang, Zijian Zhang, Sirui Ding, Piotr Mardziel, Xia Hu
• (html)Build It, Break It, Fix It: Contesting Secure Development.
  ACM Transactions on Privacy and Security (TOPS). 2020.
  James Parker, Michael Hicks, Andrew Ruef, Michelle Mazurek, Dave Levin, Daniel Votipka, Piotr Mardziel, Kelsey Fulton
• An Information-Theoretic Quantification of Discrimination with Exempt Features.
  Conference on Artificial Intelligence (AAAI). 2020.
  Sanghamitra Dutta, Praveen Venkatesh, Piotr Mardziel, Anupam Datta, Pulkit Grover
• Evaluating Design Tradeoffs in Numeric Static Analysis for Java.
  European Symposium on Programming (ESOP). 2018.
  Shiyi Wei, Piotr Mardziel, Andrew Ruef, Jeffrey S. Foster, Michael Hicks
• Quantifying vulnerability of secret generation using hyper-distributions.
  International Conference on Principles of Security and Trust (POST). 2017.
  Mario Alvim, Piotr Mardziel, Michael Hicks
• Use Privacy in Data-Driven Systems: Theory and Experiments with Machine Learnt Programs.
  Conference on Computer & Communications Security (CCS). 2017.
  Anupam Datta, Matthew Fredrikson, Gihyuk Ko, Piotr Mardziel, Shayak Sen
• Build It, Break It, Fix It: Contesting Secure Development.
  Conference on Computer & Communications Security (CCS). 2016.
  Andrew Ruef, Michael Hicks, James Parker, Dave Levin, Michelle Mazurek, Piotr Mardziel
• Towards the Quantification of Strategy Leakage.
  Workshop on Foundations of Computer Security (FCS). 2016.
  Mario Alvim, Piotr Mardziel, Michael Hicks
• Picking vs. Guessing Secrets: A Game-Theoretic Analysis.
  Computer Security Foundations Symposium (CSF). 2015.
  MHR Khouzani, Piotr Mardziel, Carlos Cid, Mudhakar Srivatsa
• Build It Break It: Measuring and Comparing Development Security.
  Workshop on Cyber Security Experimentation and Test (CSET). 2015.
  Andrew Ruef, Michael Hicks, James Parker, Dave Levin, Atif Memon, Jandelyn Plane, Piotr Mardziel
• Quantifying Information Flow for Dynamic Secrets.
  Symposium on Security and Privacy (S&P/Oakland). 2014.
  Piotr Mardziel, Mario Alvim, Michael Hicks, Michael Clarkson
• Adversary Gain vs. Defender Loss in Quantified Information Flow.
  Workshop on Foundations of Computer Security (FCS). 2014.
  Piotr Mardziel, Mario Alvim, Michael Hicks
• Dynamic Enforcement of Knowledge-based Security Policies using Probabilistic Abstract Interpretation.
  Journal of Computer Security (JCS). 2013.
  Piotr Mardziel, Stephen Magill, Michael Hicks, Mudhakar Srivatsa
• Knowledge Inference for Optimizing Secure Multi-party Computation.
  Workshop on Programming Languages and Analysis of Security (PLAS). 2013.
  Aseem Rastogi, Piotr Mardziel, Matthew Hammer, Michael Hicks
• Knowledge-Oriented Secure Multiparty Computation.
  Workshop on Programming Languages and Analysis of Security (PLAS). 2012.
  Piotr Mardziel, Michael Hicks, Jonathan Katz, Mudhakar Srivatsa
• Dynamic Enforcement of Knowledge-based Security Policies.
  Computer Security Foundations Symposium (CSF). 2011.
  Piotr Mardziel, Stephen Magill, Michael Hicks, Mudhakar Srivatsa

Tech-reports / Pre-prints

• Towards Behavior-Level Explanation for Deep Reinforcement Learning.
  arXiv. September 2020.
  Xuan Chen, Zifan Wang, Yucai Fan, Bonan Jin, Piotr Mardziel, Carlee Joe-Wong, Anupam Datta
• Supervising Feature Influence.
  arXiv. March 2018.
  Shayak Sen, Piotr Mardziel, Anupam Datta, Matthew Fredrikson
• Latent Factor Interpretations for Collaborative Filtering.
  arXiv. November 2017.
  Anupam Datta, Sophia Kovaleva, Piotr Mardziel, Shayak Sen
• Proxy Non-Discrimination in Data-Driven Systems.
  arXiv. July 2017.
  Anupam Datta, Matthew Fredrikson, Gihyuk Ko, Piotr Mardziel, Shayak Sen

Service