Official implementation of the paper "Backdoor Attacks on Self-Supervised Learning".

Overview

SSL-Backdoor

Abstract

Large-scale unlabeled data has allowed recent progress in self-supervised learning methods that learn rich visual representations. State-of-the-art self-supervised methods for learning representations from images (MoCo and BYOL) use an inductive bias that different augmentations (e.g. random crops) of an image should produce similar embeddings. We show that such methods are vulnerable to backdoor attacks where an attacker poisons a part of the unlabeled data by adding a small trigger (known to the attacker) to the images. The model performance is good on clean test images but the attacker can manipulate the decision of the model by showing the trigger at test time. Backdoor attacks have been studied extensively in supervised learning and to the best of our knowledge, we are the first to study them for self-supervised learning. Backdoor attacks are more practical in self-supervised learning since the unlabeled data is large and as a result, an inspection of the data to avoid the presence of poisoned data is prohibitive. We show that in our targeted attack, the attacker can produce many false positives for the target category by using the trigger at test time. We also develop a knowledge distillation based defense algorithm that succeeds in neutralizing the attack. Our code is available here: https://github.com/UMBCvision/SSL-Backdoor.

Paper

Backdoor Attacks on Self-Supervised Learning

Updates

  • 04/07/2021 - Poison generation code added.
  • 04/08/2021 - MoCo v2, BYOL code added.
  • 04/14/2021 - Jigsaw, RotNet code added.

Requirements

All experiments were run using the following dependencies.

  • python=3.7
  • pytorch=1.6.0
  • torchvision=0.7.0
  • wandb=0.10.21 (for BYOL)
  • torchnet=0.0.4 (for RotNet)

Optional

  • faiss=1.6.3 (for k-NN evaluation)

Create ImageNet-100 dataset

The ImageNet-100 dataset (random 100-class subset of ImageNet), commonly used in self-supervision benchmarks, was introduced in [1].

To create ImageNet-100 from ImageNet, use the provided script.

cd scripts
python create_imagenet_subset.py --subset imagenet100_classes.txt --full_imagenet_path <path> --subset_imagenet_path <path>

Poison Generation

To generate poisoned ImageNet-100 images, create your own configuration file. Some examples, which we use for our targeted attack experiments, are in the cfg directory.

  • You can choose the poisoning to be Targeted (poison only one category) or Untargeted
  • The trigger can be text or an image (We used triggers introduced in [2]).
  • The parameters of the trigger (e.g. location, size, alpha etc.) can be modified according to the experiment.
  • The poison injection rate for the training set can be modified.
  • You can choose which split to generate. "train" generates poisoned training data, "val_poisoned" poisons all the validation images for evaluation purpose. Note: The poisoned validation images are all resized and cropped to 224x224 before trigger pasting so that all poisoned images have uniform trigger size.
cd poison-generation
python generate_poison.py <configuration-file>

SSL Methods

Pytorch Custom Dataset

All images are loaded from filelists of the form given below.

<dir-name-1>/xxx.ext <target-class-index>
<dir-name-1>/xxy.ext <target-class-index>
<dir-name-1>/xxz.ext <target-class-index>

<dir-name-2>/123.ext <target-class-index>
<dir-name-2>/nsdf3.ext <target-class-index>
<dir-name-2>/asd932_.ext <target-class-index>

Evaluation

All evaluation scripts return confusion matrices for clean validation data and a csv file enumerating the TP and FP for each category.

MoCo v2 [3]

The implementation for MoCo is from https://github.com/SsnL/moco_align_uniform modified slightly to suit our experimental setup.

To train a ResNet-18 MoCo v2 model on ImageNet-100 on 2 NVIDIA GEFORCE RTX 2080 Ti GPUs:

cd moco
CUDA_VISIBLE_DEVICES=0,1 python main_moco.py \
                        -a resnet18 \
                        --lr 0.06 --batch-size 256 --multiprocessing-distributed \
                        --world-size 1 --rank 0 --aug-plus --mlp --cos --moco-align-w 0 \
                        --moco-unif-w 0 --moco-contr-w 1 --moco-contr-tau 0.2 \
                        --dist-url tcp://localhost:10005 \ 
                        --save-folder-root <path> \
                        --experiment-id <ID> <train-txt-file>

To train linear classifier on frozen MoCo v2 embeddings on ImageNet-100:

CUDA_VISIBLE_DEVICES=0 python eval_linear.py \
                        --arch moco_resnet18 \
                        --weights <SSL-model-checkpoint-path>\
                        --train_file <path> \
                        --val_file <path>

We use the linear classifier normalization from CompRess: Self-Supervised Learning by Compressing Representations which says "To reduce the computational overhead of tuning the hyperparameters per experiment, we standardize the Linear evaluation as following. We first normalize the features by L2 norm, then shift and scale each dimension to have zero mean and unit variance."

To evaluate linear classifier on clean and poisoned validation set: (This script loads the cached mean and variance from previous step.)

CUDA_VISIBLE_DEVICES=0 python eval_linear.py \
                        --arch moco_resnet18 \
                        --weights <SSL-model-checkpoint-path> \
                        --val_file <path> \
                        --val_poisoned_file <path> \
                        --resume <linear-classifier-checkpoint> \
                        --evaluate --eval_data <evaluation-ID> \
                        --load_cache

To run k-NN evaluation of frozen MoCo v2 embeddings on ImageNet-100 (faiss library needed):

CUDA_VISIBLE_DEVICES=0 python eval_knn.py \
                        -a moco_resnet18 \
                        --weights <SSL-model-checkpoint-path> \
                        --train_file <path> \
                        --val_file <path> \
                        --val_poisoned_file <path> \
                        --eval_data <evaluation-ID>

BYOL [4]

The implementation for BYOL is from https://github.com/htdt/self-supervised modified slightly to suit our experimental setup.

To train a ResNet-18 BYOL model on ImageNet-100 on 4 NVIDIA GEFORCE RTX 2080 Ti GPUs: (This scripts monitors the k-NN accuracy on clean ImageNet-100 dataset at regular intervals.)

cd byol
CUDA_VISIBLE_DEVICES=0,1,2,3 python -m train \
                                    --exp_id <ID> \
                                    --dataset imagenet --lr 2e-3 --emb 128 --method byol \
                                    --arch resnet18 --epoch 200 \
                                    --train_file_path <path> \
                                    --train_clean_file_path <path> 
                                    --val_file_path <path>
                                    --save_folder_root <path>

To train linear classifier on frozen BYOL embeddings on ImageNet-100:

CUDA_VISIBLE_DEVICES=0 python -m test --dataset imagenet \
                            --train_clean_file_path <path> \
                            --val_file_path <path> \
                            --emb 128 --method byol --arch resnet18 \
                            --fname <SSL-model-checkpoint-path>

To evaluate linear classifier on clean and poisoned validation set:

CUDA_VISIBLE_DEVICES=0 python -m test --dataset imagenet \
                            --val_file_path <path> \
                            --val_poisoned_file_path <path> \
                            --emb 128 --method byol --arch resnet18 \
                            --fname <SSL-model-checkpoint-path> \
                            --clf_chkpt <linear-classifier-checkpoint-path> \
                            --eval_data <evaluation-ID> --evaluate

Jigsaw [5]

The implementation for Jigsaw is our own Pytorch reimplementation based on the authors’ Caffe code https://github.com/MehdiNoroozi/JigsawPuzzleSolver modified slightly to suit our experimental setup. There might be some legacy Pytorch code, but that doesn't affect the correctness of training or evaluation. If you are looking for a recent Pytorch implementation of Jigsaw, https://github.com/facebookresearch/vissl is a good place to start.

To train a ResNet-18 Jigsaw model on ImageNet-100 on 1 NVIDIA GEFORCE RTX 2080 Ti GPU: (The code doesn't support Pytorch distributed training.)

cd jigsaw
CUDA_VISIBLE_DEVICES=0 python train_jigsaw.py \
                                --train_file <path> \
                                --val_file <path> \
                                --save <path>

To train linear classifier on frozen Jigsaw embeddings on ImageNet-100:

CUDA_VISIBLE_DEVICES=0 python eval_conv_linear.py \
                        -a resnet18 --train_file <path> \
                        --val_file <path> \
                        --save <path> \
                        --weights <SSL-model-checkpoint-path>

To evaluate linear classifier on clean and poisoned validation set:

CUDA_VISIBLE_DEVICES=0 python eval_conv_linear.py -a resnet18 \
                            --val_file <path> \
                            --val_poisoned_file <path> \
                            --weights <SSL-model-checkpoint-path> \
                            --resume <linear-classifier-checkpoint-path> \
                            --evaluate --eval_data <evaluation-ID>

RotNet [6]

The implementation for RotNet is from https://github.com/gidariss/FeatureLearningRotNet modified slightly to suit our experimental setup. There might be some legacy Pytorch code, but that doesn't affect the correctness of training or evaluation. If you are looking for a recent Pytorch implementation of RotNet, https://github.com/facebookresearch/vissl is a good place to start.

To train a ResNet-18 Jigsaw model on ImageNet-100 on 1 NVIDIA TITAN RTX GPU: (The code doesn't support Pytorch distributed training. Choose the experiment ID config file as required.)

cd rotnet
CUDA_VISIBLE_DEVICES=0 python main.py --exp <ImageNet100_RotNet_*> --save_folder <path>

To train linear classifier on frozen RotNet embeddings on ImageNet-100:

CUDA_VISIBLE_DEVICES=0 python main.py --exp <ImageNet100_LinearClassifiers_*> --save_folder <path>

To evaluate linear classifier on clean and poisoned validation set:

CUDA_VISIBLE_DEVICES=0 python main.py --exp <ImageNet100_LinearClassifiers_*> \
                            --save_folder <path> \
                            --evaluate --checkpoint=<epoch_num> --eval_data <evaluation-ID>

Acknowledgement

This material is based upon work partially supported by the United States Air Force under Contract No. FA8750‐19‐C‐0098, funding from SAP SE, NSF grant 1845216, and also financial assistance award number 60NANB18D279 from U.S. Department of Commerce, National Institute of Standards and Technology. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the United States Air Force, DARPA, or other funding agencies.

References

[1] Yonglong Tian, Dilip Krishnan, and Phillip Isola. Contrastive multiview coding. arXiv preprint arXiv:1906.05849,2019.

[2] Aniruddha Saha, Akshayvarun Subramanya, and Hamed Pirsiavash. Hidden trigger backdoor attacks. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, pages 11957–11965, 2020.

[3] Chen, Xinlei, et al. "Improved baselines with momentum contrastive learning." arXiv preprint arXiv:2003.04297 (2020).

[4] Jean-Bastien Grill, Florian Strub, Florent Altch́e, and et al. Bootstrap your own latent - a new approach to self-supervised learning. In Advances in Neural Information Processing Systems, volume 33, pages 21271–21284, 2020.

[5] Noroozi, Mehdi, and Paolo Favaro. "Unsupervised learning of visual representations by solving jigsaw puzzles." European conference on computer vision. Springer, Cham, 2016.

[6] Spyros Gidaris, Praveer Singh, and Nikos Komodakis. Unsupervised representation learning by predicting image rotations. In International Conference on Learning Representations, 2018.

Citation

If you find our paper, code or models useful, please cite us using

@article{saha2021backdoor,
  title={Backdoor Attacks on Self-Supervised Learning},
  author={Saha, Aniruddha and Tejankar, Ajinkya and Koohpayegani, Soroush Abbasi and Pirsiavash, Hamed},
  journal={arXiv preprint arXiv:2105.10123},
  year={2021}
}

Questions/Issues

Please create an issue on the Github Repo directly or contact [email protected] for any questions about the code.

Owner
UMBC Vision
The Computer Vision Lab at the University of Maryland, Baltimore County (UMBC)
UMBC Vision
👑 Discovery Header DoD Bug-Bounty

👑 Discovery Header DoD Bug-Bounty Did you know that DoD accepts server headers? 😲 (example: apache"version" , php"version") ? In this code it is pos

KingOfTips 38 Aug 09, 2022
CVE-2021-41773 Path Traversal for Apache 2.4.49

CVE-2021-41773 Path Traversal for Apache 2.4.49

ac1d 3 Oct 20, 2021
This is a simple Port Flooder written in Python 3.

This is a simple Port Flooder written in Python 3. Use this tool to quickly stress test your network devices and measure your router's or server's load.

Júlio Carneiro 4 Feb 20, 2022
Update of uncaptcha2 from 2019

YouTube Video Proof of Concept I created a new YouTube Video with technical Explanation for breaking Google's Audio reCAPTCHAs: Click on the image bel

Nikolai Tschacher 153 Dec 20, 2022
A python script to turn Ubuntu Desktop in a one stop security platform. The InfoSec Fortress installs the packages,tools, and resources to make Ubuntu 20.04 capable of both offensive and defensive security work.

infosec-fortress A python script to turn Ubuntu Desktop into a strong DFIR/RE System with some teeth (Purple Team Ops)! This is intended to create a s

James 41 Dec 30, 2022
CVE-2021-26855 SSRF Exchange Server

CVE-2021-26855 Brute Force EMail Exchange Server Timeline: Monday, March 8, 2021: Update Dumping content...(I'm not done, can u guy help me done this

lulz 117 Nov 28, 2022
这次是可可萝病毒!

可可萝病毒! 事情是这样的,我又开始不干正事了。 众所周知,在Python里,0x0等于0,但是不等于可可萝。 这很不好,我们得把它改成可可萝! 效果 一般的Python—— Python 3.8.0 (tags/v3.8.0:fa919fd, Oct 14 2019, 19:37:50) [MSC

黄巍 29 Jul 14, 2022
Android Malware (Analysis | Scoring) System

An Obfuscation-Neglect Android Malware Scoring System Quark-Engine is also bundled with Kali Linux, BlackArch. A trust-worthy, practical tool that's r

Quark-Engine 1k Jan 04, 2023
Password-Manager - This app can generate ,save , find and delete passwords.

Password-Manager This app can generate ,save , find and delete passwords. In the StartUp() Function , there are three buttons to choose from : Generat

1 Jan 01, 2022
Crowbar - A windows post exploitation tool

Crowbar - A windows post exploitation tool Status - ✔️ This project is now considered finished. Any updates from now on will most likely be new script

29 Nov 20, 2022
a cool, easily usable and customisable subdomains scanner

Subdah 🔎 another subdomains scanner. Installation ⚠️ Python 3.10 required ⚠️ $ git clone https://github.com/traumatism/subdah $ cd subdah $ pip3 inst

toast 14 Oct 18, 2022
Password list generator for password spraying - prebaked with goodies

Generates permutations of Months, Seasons, Years, Sports Teams (NFL, NBA, MLB, NHL), Sports Scores, "Password", and even Iterable Keyspaces of a specified size.

Casey Erdmann 65 Dec 22, 2022
PoC of proxylogon chain SSRF(CVE-2021-26855) to write file by testanull, censored by github

CVE-2021-26855 PoC of proxylogon chain SSRF(CVE-2021-26855) to write file by testanull, censored by github Why does github remove this exploit because

The Hacker's Choice 58 Nov 15, 2022
Caretaker 2 Jun 06, 2022
BurpSuite Extension: Log4j RCE Scanner

BurpSuite Extension: Log4j RCE Scanner

1 Dec 16, 2021
Click-Jack - Automatic tool to find Clickjacking Vulnerability in various Web applications

CLICK-Jack It is a automatic tool to find Clickjacking Vulnerability in various

Prince Prafull 4 Jan 10, 2022
SecurAID securely connects aid organizations directly with individuals in dangerous situations to allow them to discreetly and effectively get the assistance they need.

SecurAID securely connects aid organizations directly with individuals in dangerous situations to allow them to discreetly and effec

Ty K 2 Mar 23, 2022
This is a proof-of-concept exploit for Grafana's Unauthorized Arbitrary File Read Vulnerability (CVE-2021-43798).

CVE-2021-43798 – Grafana Exploit About This is a proof-of-concept exploit for Grafana's Unauthorized Arbitrary File Read Vulnerability (CVE-2021-43798

Pedro Havay 12 Nov 18, 2022
TCP/UDP port scanner on python, usong scapy and multiprocessin

Port Scanner TCP/UDP port scanner on python, usong scapy and multiprocessing. Usage python3 scanner.py [OPTIONS] IP_ADDRESS [{tcp|udp}[/[PORT|PORT-POR

Egor Krokhin 1 Dec 05, 2021
This project is for finding a solution to use Security Onion Elastic data with Jupyter Notebooks.

This project is for finding a solution to use Security Onion Elastic data with Jupyter Notebooks. The goal is to successfully use this notebook project below with Security Onion for beacon detection

4 Jun 08, 2022