Global Triggers for Attacking and Analyzing Ranking Models

Show simple item record

dc.identifier.uri Wang, Yumeng eng
dc.contributor.other L3S Research Center 2022-07-18T10:41:34Z 2022-07-18T10:41:34Z 2022
dc.identifier.citation Wang, Yumeng: Global Triggers for Attacking and Analyzing Ranking Models. Hannover : Gottfried Wilhelm Leibniz Universität Hannover, Institut für Verteilte Systeme, Master Thesis, 2022, VII, 70 S. DOI: eng
dc.description.abstract Text ranking models based on BERT are now well established for a wide range of pas- sage and document ranking tasks. However, the robustness of BERT-based ranking models under adversarial attack is under-explored. In this work, we argue that BERT- rankers are vulnerable to adversarial attacks targeting retrieved documents given a query. We propose algorithms for generating adversarial perturbation of documents locally to individual queries or globally across the dataset using gradient-based optimization methods. The aim of our algorithms is to add a small number of tokens to a highly relevant or non-relevant document to cause a significant rank demotion or promotion. Our experiments show that a few number of tokens can already change the document rank by a large margin. Besides, we find that BERT-rankers heavily rely on the docu- ment start/head for relevance prediction, making the initial part of the document more susceptible to adversarial attacks. More interestingly, our statistical analysis finds a small set of recurring adversar- ial tokens that when concatenated to documents result in successful rank demo- tion/promotion of any relevant/non-relevant document respectively. Finally, our ad- versarial tokens also show particular topic preferences within and across datasets, exposing potential biases from BERT pre-training or downstream datasets. eng
dc.language.iso eng eng
dc.publisher Hannover : Gottfried Wilhelm Leibniz Universität
dc.rights CC BY 3.0 DE eng
dc.rights.uri eng
dc.subject ranking eng
dc.subject BERT eng
dc.subject neural networks eng
dc.subject neural network eng
dc.subject biases eng
dc.subject.ddc 004 | Informatik eng
dc.title Global Triggers for Attacking and Analyzing Ranking Models eng
dc.type MasterThesis eng
dc.type Text eng
dcterms.extent VII, 70 S.
dc.description.version publishedVersion eng
tib.accessRights frei zug�nglich eng

Files in this item

This item appears in the following Collection(s):

Show simple item record


Search the repository


My Account

Usage Statistics