Built independently by an author, for readers. Read the story and support ChapterPal
ChapterPal Logo
Log inSign up

keyword

backdoor defense

Backdoor defense refers to a set of security techniques and countermeasures designed to detect, prevent, or eliminate hidden malicious vulnerabilities within machine learning models and datasets. In a backdoor attack, an adversary manipulates a model during training or development so that it performs accurately on typical inputs but outputs a compromised, targeted prediction whenever a specific trigger pattern is present. Backdoor defenses mitigate these risks throughout the machine learning lifecycle by filtering poisoned training samples, employing robust training mechanisms, auditing model parameters, and repairing infected networks through methods such as neuron pruning, fine-tuning, or trigger unlearning. Additionally, these defenses can operate during inference to identify and sanitize triggered inputs, ensuring that the system remains secure without sacrificing its predictive performance on benign data.

1 item

Reconstructive Neuron Pruning for Backdoor Defense

Reconstructive Neuron Pruning for Backdoor Defense

Yige Li, Xixiang Lyu, Xingjun Ma, Nodens Koren, Lingjuan Lyu, Bo Li, Yu-Gang Jiang

OrganizationsFudan UniversityShanghai Artificial Intelligence LaboratorySony CorporationUniversity of CopenhagenUniversity of Illinois Urbana-ChampaignXidian University

Why you should read this

Proposes an asymmetric unlearning and filter-recovery framework that exposes and prunes backdoor neurons using only a small set of clean data, effectively purifying backdoored models across diverse attacks without sacrificing clean classification accuracy.

Deep neural networks (DNNs) have been found to be vulnerable to backdoor attacks, raising security concerns about their deployment in mission-critical applications. While existing defense methods have demonstrated promising results, it is still not clear how to effectively remove backdoor-associated neurons in backdoored DNNs. In this paper, we propose a novel defense called Reconstructive Neuron Pruning (RNP) to expose and prune backdoor neurons via an unlearning and then recovering process. Specifically, RNP first unlearns the neurons by maximizing the model’s error on a small subset of clean samples and then recovers the neurons by minimizing the model’s error on the same data. In RNP, unlearning is operated at the neuron level while recovering is operated at the filter level, forming an asymmetric reconstructive learning procedure. We show that such an asymmetric process on only a few clean samples can effectively expose and prune the backdoor neurons implanted by a wide range of attacks, achieving a new state-of-the-art defense performance. Moreover, the unlearned model at the intermediate step of our RNP can be directly used to improve other backdoor defense tasks including backdoor removal, trigger recovery, backdoor label detection, and backdoor sample detection. Code is available at https://github.com/bboylyg/RNP.

Added

2026-09-26