• Home
  • Search
  • Detection of Backdoors in Trained Classifiers Without Access to the Training Set.
  • Open Access IconOpen Access
  • Cite Icon51
  • https://doi.org/10.1109/tnnls.2020.3041202Copy DOI Icon

Detection of Backdoors in Trained Classifiers Without Access to the Training Set.

Show More
  • Abstract
  • Literature Map
  • References
  • Citations
  • Similar Papers
Abstract

With wide deployment of deep neural network (DNN) classifiers, there is great potential for harm from adversarial learning attacks. Recently, a special type of data poisoning (DP) attack, known as a backdoor (or Trojan), was proposed. These attacks do not seek to degrade classification accuracy, but rather to have the classifier learn to classify to a target class t∗ whenever the backdoor pattern is present in a test example originally from a source class s∗ . Launching backdoor attacks does not require knowledge of the classifier or its training process-only the ability to poison the training set with exemplars containing a backdoor pattern (labeled with the target class). Defenses against backdoors can be deployed before/during training, post-training, or at test time. Here, we address post-training detection in DNN image classifiers, seldom considered in existing works, wherein the defender does not have access to the poisoned training set, but only to the trained classifier itself, as well as to clean (unpoisoned) examples from the classification domain. This scenario is of great interest because e.g., a classifier may be the basis of a phone app that will be shared with many users. Detection may thus reveal a widespread attack. We propose a purely unsupervised anomaly detection (AD) defense against imperceptible backdoor attacks that: 1) detects whether the trained DNN has been backdoor-attacked; 2) infers the source and target classes in a detected attack; 3) estimates the backdoor pattern itself. Our AD approach involves learning (via suitable cost function minimization) the minimum size/norm perturbation (putative backdoor) required to induce the classifier to misclassify (most) examples from class s to class t , for all (s,t) pairs. Our hypothesis is that nonattacked pairs require large perturbations, while the attacked pair (s∗, t∗) requires much smaller ones. This is convincingly borne out experimentally. We identify a variety of plausible cost functions and devise a novel, robust hypothesis testing approach to perform detection inference. We test our approach, in comparison with the state-of-the-art methods, for several backdoor patterns, attack settings and mechanisms, and data sets and demonstrate its favorability. Our defense essentially requires setting a single hyperparameter (the detection threshold), which can e.g., be chosen to fix the system's false positive rate.

Similar Papers
  • Conference Article
  • Citations43

Revealing Backdoors, Post-Training, in DNN Classifiers via Novel Inference on Optimized Perturbations Inducing Group Misclassification

  • Apr 02, 2020
  • Zhen Xiang +2
  • Conference Article
  • Citations10

Test-Time Detection of Backdoor Triggers for Poisoned Deep Neural Networks

  • May 23, 2022
  • Xi Li +3
  • Conference Article
  • Citations28

Vulnerabilities of Deep Learning-Driven Semantic Communications to Backdoor (Trojan) Attacks

  • Mar 22, 2023
  • Yalin E Sagduyu +3
  • Research Article
  • Citations12

Detecting backdoor in deep neural networks via intentional adversarial perturbations

  • Mar 22, 2023
  • Information Sciences
  • Mingfu Xue +5
  • Research Article
  • Citations25

Reverse engineering imperceptible backdoor attacks on deep neural networks for detection and training set cleansing

  • Apr 22, 2021
  • Computers & Security
  • Zhen Xiang +2
  • Conference Article
  • Citations238

Clean-Label Backdoor Attacks on Video Recognition Models

  • Jun 01, 2020
  • Shihao Zhao +5
  • Conference Article
  • Citations12

L-Red: Efficient Post-Training Detection of Imperceptible Backdoor Attacks Without Access to the Training Set

  • Jun 06, 2021
  • Zhen Xiang +2
  • Research Article
  • Citations6

Adversarial-Inspired Backdoor Defense via Bridging Backdoor and Adversarial Attacks

  • Apr 11, 2025
  • Proceedings of the AAAI Conference on Artificial Intelligence
  • Jia-Li Yin +4
  • Research Article
  • Citations45

When Not to Classify: Anomaly Detection of Attacks (ADA) on DNN Classifiers at Test Time.

  • Jul 01, 2019
  • Neural Computation
  • David Miller +2
  • Conference Article
  • Citations16

Towards Backdoor Attack on Deep Learning based Time Series Classification

  • May 01, 2022
  • Daizong Ding +6
  • Conference Article
  • Citations8

Detecting Backdoor Attacks against Point Cloud Classifiers

  • May 23, 2022
  • Zhen Xiang +4
  • Research Article
  • Citations11

Backdoor Federated Learning-Based mmWave Beam Selection

  • Oct 01, 2022
  • IEEE Transactions on Communications
  • Zhengming Zhang +5
  • Research Article
  • Citations26

Backdoor Attacks on Image Classification Models in Deep Neural Networks

  • Mar 01, 2022
  • Chinese Journal of Electronics
  • Quanxin Zhang +5
  • Research Article
  • Citations34

PTB: Robust physical backdoor attacks against deep neural networks in real world

  • Apr 15, 2022
  • Computers & Security
  • Mingfu Xue +6
  • Conference Article
  • Citations45

A Benchmark Study Of Backdoor Data Poisoning Defenses For Deep Neural Network Classifiers And A Novel Defense

  • Oct 01, 2019
  • Zhen Xiang +2
Cactus Communications logo

Copyright 2026 Cactus Communications. All rights reserved.