analiza aiqbench
This research addresses the emerging threat of LLM agents generating coordinated shilling attacks on recommender systems, proposing TH-GNN as a novel detection framework. By jointly modeling temporal patterns, graph structure, and semantic content from frozen language models, the approach significantly outperforms existing text-only or graph-only baselines in identifying these sophisticated cross-modal inconsistencies. The work highlights a critical intersection between advanced machine learning architectures and the security challenges posed by autonomous AI agents.
TH-GNN: Heterogeneous Temporal Graph Neural Networks for LLM-Agent Shilling Attack Detection
arXiv:2608.20376v1 Announce Type: cross Abstract: LLM agents can now generate realistic shilling profiles, fluent reviews, and coherent ratings at scale, systematically defeating recommender-system defenses. Text-only detectors that flag semantic drift in review embeddings are blind to graph structure and temporal coordination, while graph-only detectors that exploit neighborhood anomalies cannot reason over review semantics or the cross-modal inconsistencies produced by LLM-generated content. We propose TH-GNN, a heterogeneous temporal graph neural network with a two-layer Heterogeneous Graph Transformer backbone that applies per-type and per-relation attention augmented with learnable sinusoidal temporal encodings on every edge. Cross-modal attention fuses structural user embeddings with frozen RoBERTa representations of reviews and item descriptions, while a GRU operating over log inter-arrival times captures temporal burstiness. Evaluated across five attack families and four benchmark datasets, TH-GNN achieves a grand-mean F1 score of 0.870, outperforming the strongest text-only baseline on Agent4SR attacks by 10.9 percentage points and 11.5 percentage points at the lowest injection rate. These results demonstrate the effectiveness of jointly modeling temporal, structural, and semantic signals for detecting sophisticated LLM-driven shilling attacks.
Abstract:LLM agents can now generate realistic shilling profiles, fluent reviews, and coherent ratings at scale, systematically defeating recommender-system defenses. Text-only detectors that flag semantic drift in review embeddings are blind to graph structure and temporal coordination, while graph-only detectors that exploit neighborhood anomalies cannot reason over review semantics or the cross-modal inconsistencies produced by LLM-generated content. We propose TH-GNN, a heterogeneous temporal graph neural network with a two-layer Heterogeneous Graph Transformer backbone that applies per-type and per-relation attention augmented with learnable sinusoidal temporal encodings on every edge. Cross-modal attention fuses structural user embeddings with frozen RoBERTa representations of reviews and item descriptions, while a GRU operating over log inter-arrival times captures temporal burstiness. Evaluated across five attack families and four benchmark datasets, TH-GNN achieves a grand-mean F1 score of 0.870, outperforming the strongest text-only baseline on Agent4SR attacks by 10.9 percentage points and 11.5 percentage points at the lowest injection rate. These results demonstrate the effectiveness of jointly modeling temporal, structural, and semantic signals for detecting sophisticated LLM-driven shilling attacks.
Computation and Language (cs.CL); Machine Learning (cs.LG)
arXiv:2608.20376 [cs.CL]
(or arXiv:2608.20376v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2608.20376 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 266 cuvinte
Subiecte conexe
CAT: Can Trust be Predicted with Context-Awareness in Dynamic Heterogeneous Networks?
arXiv · 30 sept. 2026
HERMES: Heterogeneous Edge-Relational Multi-Head Embedded SSM Attention for Traffic Conflict Prediction at Signalized Intersections
arXiv · 24 iul. 2026
A Survey of Heterogeneous Graph Neural Networks for Cybersecurity Anomaly Detection
arXiv · 9 iun. 2026
Links #5: 2026/07
LessWrong · 31 iul. 2026