Skip to content
Book Open access

From Auxiliary Views to Actions: Edge Augmentation for Strict Cold-Start Fraud Screening

Nov 2026 · Proceedings of the 35th ACM International Conference on Information and Knowledge Management · pp. 4854-4864 · 0 citations · 55 references

TL;DR

A novel graph learning framework called CoMGL is proposed that employs a multi-view learning strategy to effectively tackle SCS tasks on graphs and regards the interaction bipartite graphs from different scenarios as the multi-views for separately learning user representation, yielding more comprehensive representation for SCS users.

Abstract

Fraudulent user detection in advance is crucial for risk control of real-world applications. One of the primary challenges lies in classifying new users as either benign or fraudulent when only limited user profile information is available. This scenario is often referred to as the Strict Cold Start (SCS) problem. Existing methods generally address this challenge by enhancing the topology or features of SCS nodes. However, these approaches can introduce task-irrelevant noise or require sharing raw cross-scenario features, which is operationally undesirable when scenarios are owned by different teams. To overcome these limitations, we propose a novel graph learning framework called CoMGL. This framework employs a multi-view learning strategy to effectively tackle SCS tasks on graphs. Specifically, the CoMGL regards the interaction bipartite graphs from different scenarios as the multi-views for separately learning user representation, yielding more comprehensive representation for SCS users. Each view encoder consumes only its own scenario-local features, and only the resulting user embeddings are fused, which limits cross-scenario raw feature transfer. Experimental results on real-world datasets demonstrate the effectiveness of the proposed framework compared with the state-of-the-art approaches.

Read PDF

Similar papers

Book Open access Nov 2026

An Unseen Threat in Cross-Domain Recommendation: User Re-Identification Attacks and Practical Defenses

Cross-domain recommendation (CDR) boosts personalization by transferring knowledge across domains, but its privacy implications remain largely overlooked. In this paper, we uncover a critical threat in CDR: cross-domain user re-identification can be performed in realistic black-box settings using only the top-K recomme...

Tian-Zi Zang, Shuai-Jie Zhou, Yi-Cong Li et al. · 0 citations
Preprint Oct 2026

Label-Budgeted Attribute Inference from Protected Features in Edge-Cloud Split Multimodal LLMs

Edge-cloud split inference runs the early layers of a multimodal large language model (MLLM) on a device and sends intermediate features to a server to complete the task. While keeping raw images local is often perceived as a privacy benefit, an honest-but-curious server can still infer sensitive attributes, such as ag...

Jung-Chun Yao, Peichun Hua, Han-Xiu Zhang et al. · 0 citations

Beyond Accuracy: Centroid-Guided Contrastive Loss for Structured Fraudulent Job Posting Detection

Centroid-Guided Contrastive Loss (CGCL), a loss function which unifies classification with densely formulated clustering to consistently reshape latent-space through a centroid-driven top-$k$ push-and-pull mechanism, is proposed.

Syed Ali Ahmed, M. Raza, Muhammad Shoaib Siddiqui et al. · 0 citations
Preprint Oct 2026

CARTA: Self-Supervised Reusable Transaction Embeddings via Heterogeneous Tokenization and Cross-Attention

Acquiring fraud labels for payment transactions is expensive and inconsistent across tasks, and high-cardinality identifiers such as merchant names make one-hot feature pipelines infeasible at scale. We present CARTA, a self-supervised encoder that learns a compact 16-dim embedding from raw transactions without labels;...

Shreyas K. Chandrahas, Jose M. Munoz, Satish Babu Bellapu · 0 citations
#artificial intelligence Preprint Oct 2026

Feature-Context Consistency: Unsupervised Adversarial Detection with Drift Stability on Attributed Graphs

Adversarial detection on evolving attributed graphs faces two challenges: fraud labels arrive late or not at all, and natural distributional drift erodes any trained decision boundary. Modern detectors can reach strong detection rates when labels are available and the distribution is stationary; sustaining that quality...

Firas Bayram, Maxime Cordy · 0 citations
Preprint Oct 2026

Embedding Asymmetry as Attack Signal: Detecting and Defending Active Label Inference in Vertical Federated Learning

Vertical federated learning lets organizations train a shared model over features they hold separately, while one party keeps the labels private. The active label inference attack of Fu et al. [12] breaks this: a feature-holding party runs a modified optimizer, MaliciousSGD, that forces class structure into its embeddi...

Soumya Swagata Biswas, M. Adnan · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.