What Research Areas Does the SAND Lab Focus On?

Updated

The SAND Lab (Security, Algorithms, Networking and Data) at the University of Chicago centers on security and machine learning, with occasional work in HCI and data mining. Its listed research areas include Adversarial Machine Learning and Wireless and Mobile Systems. If your interests sit at the intersection of security and ML — especially attacks on or defenses for generative models — this lab's direction is likely a strong match.

Core Research Areas

The lab describes its work as spanning "topics in security and machine learning, with occasional ventures into HCI and data mining." The site's own headings name two current research areas explicitly:

  • Adversarial Machine Learning — attacks and defenses involving ML models, including data poisoning and detection.
  • Wireless and Mobile Systems — systems-level security and networking work.

Alongside these, the lab's recent output points to a heavy emphasis on generative AI security and privacy, which sits within the adversarial ML umbrella.

What the Recent Work Shows

The "Latest" section is the clearest signal of where the lab's attention currently goes. Recurring themes include:

Theme Example work mentioned
Data poisoning / adversarial mislabeling Hemlock (adversarial mislabeling poison) attacks, CCS
Generative AI image detection Generative AI image detection paper, CCS Distinguished Paper Award
Model manipulation Nightshade (IEEE S&P 2024); diffusion model implosion (ACM CCS 2024)
Privacy attacks AI-enhanced privacy attacks (PhD dissertation topic)
Measurement / crawling Measuring genAI crawlers and defenses (IMC)
Model versioning Model versioning paper (SaTML 2024)

A few observations from this list:

  • Poisoning and detection are paired concerns. The lab works on both attacks (e.g., mislabeling poisons, Nightshade) and defenses/detection (e.g., generative AI image detection), so it isn't purely offense- or defense-oriented.
  • Generative models are the current focal point. Much of the recent work targets diffusion models and generative AI rather than classical ML.
  • Measurement work appears too, such as crawling and measuring genAI crawlers, which fits the "Networks and Data" part of the lab's name.

Who Runs It and Who's In It

The lab is co-directed by Professor Ben Y. Zhao and Professor Heather Zheng. Members include roughly 6–10 PhD, MS, and undergraduate researchers, plus visiting scholars. If you're evaluating fit, note that the lab is co-led, so prospective students are typically advised by one or both directors.

How to Judge Whether It Matches Your Interests

Use these conditions to decide:

  • Strong match if you want to work on adversarial ML, generative AI security/privacy, or data poisoning and detection.
  • Possible match if your interest is wireless/mobile systems security, since that remains a listed area.
  • Weaker match if your primary interest is HCI or data mining — the lab describes these as occasional ventures rather than core areas.

For prospective PhD applicants, the site notes the lab is looking to recruit 1–2 students in the 2025–2026 admission cycle, asks interested candidates to fill out a Google form to stay on a short list, and advises mentioning Heather and Ben when selecting faculty in the UChicago PhD application.