Summary of Few-shot Model Extraction Attacks Against Sequential Recommender Systems, by Hui Zhang et al.
Few-shot Model Extraction Attacks against Sequential Recommender Systems
by Hui Zhang, Fu Liu
First submitted to arxiv on: 18 Nov 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Cryptography and Security (cs.CR); Information Retrieval (cs.IR)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes a novel few-shot model extraction framework against sequential recommenders, aiming to construct a superior surrogate model using limited data (10% or less). The approach combines an autoregressive augmentation generation strategy and a bidirectional repair loss-facilitated model distillation procedure. The former generates synthetic data by extracting inherent dependencies and characterizing user behavioral patterns, while the latter rectifies errors in surrogate models through auxiliary losses that target recommendation list discrepancies. Experimental results on three datasets demonstrate the effectiveness of this framework. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper solves a big problem in keeping our favorite apps from recommending weird stuff to us. Right now, there’s a way to trick these apps into suggesting things we don’t like just by looking at how they work. But what if someone only had a tiny bit of data? That would make it even harder! The researchers came up with a new plan to make a fake version of the app that works almost as well as the real one, using very little information. They tested this idea on three different apps and showed it really works! |
Keywords
» Artificial intelligence » Autoregressive » Distillation » Few shot » Synthetic data
