Investigating the Impact of Pre-trained Word Embeddings on Memorization in Neural Networks - Archive ouverte HAL Access content directly
Conference Papers Year : 2020

Investigating the Impact of Pre-trained Word Embeddings on Memorization in Neural Networks

(1) , (1) , (1) , (1) , (1)
1

Abstract

The sensitive information present in the training data, poses a privacy concern for applications as their unintended memorization during training can make models susceptible to membership inference and attribute inference attacks. In this paper, we investigate this problem in various pre-trained word embeddings (GloVe, ELMo and BERT) with the help of language models built on top of it. In particular, firstly sequences containing sensitive information like a single-word disease and 4-digit PIN are randomly inserted into the training data, then a language model is trained using word vectors as input features, and memorization is measured with a metric termed as exposure. The embedding dimension , the number of training epochs, and the length of the secret information were observed to affect memorization in pre-trained embeddings. Finally, to address the problem, differentially private language models were trained to reduce the exposure of sensitive information.
Fichier principal
Vignette du fichier
ThomasA+20.pdf (347.1 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-02880590 , version 1 (25-06-2020)

Identifiers

  • HAL Id : hal-02880590 , version 1

Cite

Aleena Thomas, David Ifeoluwa Adelani, Ali Davody, Aditya Mogadala, Dietrich Klakow. Investigating the Impact of Pre-trained Word Embeddings on Memorization in Neural Networks. 23rd International Conference on Text, Speech and Dialogue, Sep 2020, brno, Czech Republic. ⟨hal-02880590⟩
135 View
920 Download

Share

Gmail Facebook Twitter LinkedIn More