Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The Role of Fake Users in Sequential Recommender Systems

Filippo Betello

TL;DR

The paper addresses the robustness of Sequential Recommender Systems to training-time fake-user perturbations by evaluating two architectures, SASRec and GRU4Rec, on four real-world datasets with four fake-user types and varying proportions. While standard metrics like $NDCG$ show limited sensitivity, Rank List Sensitivity (RLS) reveals substantial shifts in the produced recommendation lists, indicating degraded ranking stability under perturbations. Key findings include dataset- and model-dependent variations, notably near-zero Jaccard overlap in some cases and underrepresentation of unpopular items in top-$k$ lists, suggesting that attackers could manipulate perceived recommendations without dramatically changing traditional scores. The work highlights the need for robust SRS training strategies and outlines future directions for resilience, broader evaluation, and category-focused analyses to mitigate adversarial effects in sequential recommendations.

Abstract

Sequential Recommender Systems (SRSs) are widely used to model user behavior over time, yet their robustness remains an under-explored area of research. In this paper, we conduct an empirical study to assess how the presence of fake users, who engage in random interactions, follow popular or unpopular items, or focus on a single genre, impacts the performance of SRSs in real-world scenarios. We evaluate two SRS models across multiple datasets, using established metrics such as Normalized Discounted Cumulative Gain (NDCG) and Rank Sensitivity List (RLS) to measure performance. While traditional metrics like NDCG remain relatively stable, our findings reveal that the presence of fake users severely degrades RLS metrics, often reducing them to near-zero values. These results highlight the need for further investigation into the effects of fake users on training data and emphasize the importance of developing more resilient SRSs that can withstand different types of adversarial attacks.

The Role of Fake Users in Sequential Recommender Systems

TL;DR

The paper addresses the robustness of Sequential Recommender Systems to training-time fake-user perturbations by evaluating two architectures, SASRec and GRU4Rec, on four real-world datasets with four fake-user types and varying proportions. While standard metrics like show limited sensitivity, Rank List Sensitivity (RLS) reveals substantial shifts in the produced recommendation lists, indicating degraded ranking stability under perturbations. Key findings include dataset- and model-dependent variations, notably near-zero Jaccard overlap in some cases and underrepresentation of unpopular items in top- lists, suggesting that attackers could manipulate perceived recommendations without dramatically changing traditional scores. The work highlights the need for robust SRS training strategies and outlines future directions for resilience, broader evaluation, and category-focused analyses to mitigate adversarial effects in sequential recommendations.

Abstract

Sequential Recommender Systems (SRSs) are widely used to model user behavior over time, yet their robustness remains an under-explored area of research. In this paper, we conduct an empirical study to assess how the presence of fake users, who engage in random interactions, follow popular or unpopular items, or focus on a single genre, impacts the performance of SRSs in real-world scenarios. We evaluate two SRS models across multiple datasets, using established metrics such as Normalized Discounted Cumulative Gain (NDCG) and Rank Sensitivity List (RLS) to measure performance. While traditional metrics like NDCG remain relatively stable, our findings reveal that the presence of fake users severely degrades RLS metrics, often reducing them to near-zero values. These results highlight the need for further investigation into the effects of fake users on training data and emphasize the importance of developing more resilient SRSs that can withstand different types of adversarial attacks.

Paper Structure

This paper contains 16 sections, 1 equation, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Sequential Recommender Systems
  4. Training Perturbations
  5. Methodology
  6. Background
  7. Fake user design
  8. Models
  9. Datasets
  10. Evaluation
  11. Experimental Setup
  12. Results
  13. RQ1: Impact of Fake Users on Standard Metrics for Real Users
  14. RQ2: Analysis of Recommendation Lists Generated for Real Users
  15. RQ3: Influence of Fake User Interactions on Popular and Unpopular Items
  16. ...and 1 more sections

Figures (2)

  • Figure 1: Plots of various metrics for all the datasets considered as the percentage of fake users increases. The baseline is shown as a horizontal solid line, while other lines show the metrics as the percentage of fake users changes for the three scenarios considered.
  • Figure 2: Plots of RLS metrics for all the datasets considered as the percentage of fake users increases. The metrics are shown as the percentage of fake users changes for the three scenarios considered.