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

keyword

adaptive frequency trigger

An adaptive frequency trigger is a stealthy backdoor mechanism used against deep learning models, where imperceptible trigger patterns are dynamically generated and injected into the frequency domain of input data rather than its spatial representation. In this technique, transformations such as the discrete cosine transform are applied to input signals or images, allowing the attack to adaptively alter specific frequency coefficients based on the input content and targeted attack objectives. By operating within the frequency domain, the trigger remains resistant to standard image compression and processing defenses while preserving visual fidelity to human observers. When presented to a compromised neural network, the modified frequency components activate pre-programmed malicious behaviors, such as degrading data compression efficiency, compromising reconstruction quality, or manipulating outputs in downstream vision tasks like classification and semantic segmentation.

1 item

Backdoor Attacks Against Deep Image Compression via Adaptive Frequency Trigger

Backdoor Attacks Against Deep Image Compression via Adaptive Frequency Trigger

Yi Yu, Yufei Wang, Wenhan Yang, Shijian Lu, Yap-Peng Tan, Alex C. Kot

OrganizationsNanyang Technological UniversityPeng Cheng Laboratory

Why you should read this

Presents a frequency-based backdoor attack targeting deep image compression models by injecting adaptive discrete cosine transform triggers into only the encoder, effectively compromising reconstruction quality, bit-rate, and downstream vision tasks without altering the decoder.

Recent deep-learning-based compression methods have achieved superior performance compared with traditional approaches. However, deep learning models have proven to be vulnerable to backdoor attacks, where some specific trigger patterns added to the input can lead to malicious behavior of the models. In this paper, we present a novel backdoor attack with multiple triggers against learned image compression models. Motivated by the widely used discrete cosine transform (DCT) in existing compression systems and standards, we propose a frequency-based trigger injection model that adds triggers in the DCT domain. In particular, we design several attack objectives for various attacking scenarios, including: 1) attacking compression quality in terms of bit-rate and reconstruction quality; 2) attacking task-driven measures, such as down-stream face recognition and semantic segmentation. Moreover, a novel simple dynamic loss is designed to balance the influence of different loss terms adaptively, which helps achieve more efficient training. Extensive experiments show that with our trained trigger injection models and simple modification of encoder parameters (of the compression model), the proposed attack can successfully inject several backdoors with corresponding triggers in a single image compression model.

Added

2026-09-26