GAP: Differentially Private Graph Neural Networks with Aggregation Perturbation

Sajadmanesh, Sina; Shamsabadi, Ali Shahin; Bellet, Aurelien; Gatica-Perez, Daniel

Type de document :

Communication dans un congrès avec actes

Titre :

GAP: Differentially Private Graph Neural Networks with Aggregation Perturbation

Auteur(s) :

Sajadmanesh, Sina [Auteur]
Ecole Polytechnique Fédérale de Lausanne [EPFL]
IDIAP Research Institute
Shamsabadi, Ali Shahin [Auteur]
The Alan Turing Institute
Bellet, Aurelien [Auteur]

Machine Learning in Information Networks [MAGNET]
Gatica-Perez, Daniel [Auteur]
Ecole Polytechnique Fédérale de Lausanne [EPFL]
IDIAP Research Institute

Titre de la manifestation scientifique :

USENIX Security 2023 - 32nd USENIX Security Symposium

Ville :

Anaheim

Pays :

Etats-Unis d'Amérique

Date de début de la manifestation scientifique :

2023-08-09

Discipline(s) HAL :

Informatique [cs]/Apprentissage [cs.LG]
Statistiques [stat]/Machine Learning [stat.ML]

Résumé en anglais : [en]

In this paper, we study the problem of learning Graph Neural Networks (GNNs) with Differential Privacy (DP). We propose a novel differentially private GNN based on Aggregation Perturbation (GAP), which adds stochastic noise ...
Lire la suite >In this paper, we study the problem of learning Graph Neural Networks (GNNs) with Differential Privacy (DP). We propose a novel differentially private GNN based on Aggregation Perturbation (GAP), which adds stochastic noise to the GNN's aggregation function to statistically obfuscate the presence of a single edge (edge-level privacy) or a single node and all its adjacent edges (node-level privacy). Tailored to the specifics of private learning, GAP's new architecture is composed of three separate modules: (i) the encoder module, where we learn private node embeddings without relying on the edge information; (ii) the aggregation module, where we compute noisy aggregated node embeddings based on the graph structure; and (iii) the classification module, where we train a neural network on the private aggregations for node classification without further querying the graph edges. GAP's major advantage over previous approaches is that it can benefit from multi-hop neighborhood aggregations, and guarantees both edge-level and node-level DP not only for training, but also at inference with no additional costs beyond the training's privacy budget. We analyze GAP's formal privacy guarantees using Rényi DP and conduct empirical experiments over three real-world graph datasets. We demonstrate that GAP offers significantly better accuracy-privacy trade-offs than state-of-the-art DP-GNN approaches and naive MLP-based baselines. Our code is publicly available at https://github.com/sisaman/GAP.Lire moins >

Langue :

Anglais

Comité de lecture :

Oui

Audience :

Internationale

Vulgarisation :

Non

Projet ANR :

Apprentissage automatique décentralisé et préservant la vie privée

Collections :

Centre de Recherche en Informatique, Signal et Automatique de Lille (CRIStAL) - UMR 9189

Source :

Harvested from HAL

Fichiers

document
Accès libre
Accéder au document

2203.00949.pdf
Accès libre
Accéder au document

2203.00949
Accès libre
Accéder au document

document
Accès libre
Accéder au document

2203.00949.pdf
Accès libre
Accéder au document

GAP: Differentially Private Graph Neural ... BibTeX CSV Excel RIS

Fichiers

GAP: Differentially Private Graph Neural ...

BibTeX

CSV

Excel

RIS