BadFusion: 2D-Oriented Backdoor Attacks against 3D Object Detection

Authors: Saket S. Chaturvedi, Lan Zhang, Wenbin Zhang, Pan He, Xiaoyong Yuan

IJCAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 5 Evaluation In this section, we first detail our experimental framework (dataset, implementation & training details, evaluation metrics) and then present the evaluation results of the proposed Bad Fusion. We further demonstrate the effectiveness of Bad Fusion against mainly Point-line Camera-to-Li DAR fusion methods in 3D object detection and also benchmark our approach against three state-of-the-art backdoor detection methods. Lastly, we conduct an ablation study to elucidate the internal mechanics of the Bad Fusion.
Researcher Affiliation Academia 1Clemson University 2Florida International University 3Auburn University
Pseudocode Yes Algorithm 1 Algorithm Procedure of Bad Fusion
Open Source Code No The paper states: 'We implement Untar OD based on their open-source code3' with a link to a third-party repository. There is no explicit statement or link indicating that the code for their proposed method (Bad Fusion) is open-source or publicly available.
Open Datasets Yes We use the KITTI dataset [Geiger et al., 2013] in the evaluation.
Dataset Splits Yes we split the training data into a train set and a validation set with 3, 712 and 3, 769 samples, respectively, following the train/valid split process in previous work [Chen et al., 2016].
Hardware Specification No The paper does not explicitly specify the hardware used for running the experiments (e.g., GPU/CPU models, memory specifications).
Software Dependencies No The paper mentions software components like 'Focal Loss', 'Smooth L1Loss', 'Adam W optimizer', and 'mmdetection3d' but does not provide specific version numbers for these or other key software dependencies required for replication.
Experiment Setup Yes The fusion models are trained using an Adam W optimizer with a learning rate of 0.002 and a weight decay parameter of 0.01 for 70 epochs.