Phantom: Untargeted Poisoning Attacks on Semi-Supervised Learning (Full Version)

TL;DR

This paper introduces Phantom, a novel untargeted poisoning attack on semi-supervised learning that manipulates unlabeled data to significantly degrade model performance, highlighting security vulnerabilities in user-generated content platforms.

Contribution

We propose Phantom, the first untargeted poisoning attack on SSL that requires only minimal manipulated samples, demonstrating its effectiveness across multiple datasets and social media platforms.

Findings

Small fractions of manipulated data reduce accuracy by 10%

Phantom is effective across 6 datasets and 3 social media platforms

High manipulation levels can degrade SSL models to naive classifiers

Abstract

Deep Neural Networks (DNNs) can handle increasingly complex tasks, albeit they require rapidly expanding training datasets. Collecting data from platforms with user-generated content, such as social networks, has significantly eased the acquisition of large datasets for training DNNs. Despite these advancements, the manual labeling process remains a substantial challenge in terms of both time and cost. In response, Semi-Supervised Learning (SSL) approaches have emerged, where only a small fraction of the dataset needs to be labeled, leaving the majority unlabeled. However, leveraging data from untrusted sources like social networks also creates new security risks, as potential attackers can easily inject manipulated samples. Previous research on the security of SSL primarily focused on injecting backdoors into trained models, while less attention was given to the more challenging untargeted poisoning attacks. In this paper, we introduce Phantom, the first untargeted poisoning attack in SSL that disrupts the training process by injecting a small number of manipulated images into the unlabeled dataset. Unlike existing attacks, our approach only requires adding few manipulated samples, such as posting images on social networks, without the need to control the victim. Phantom causes SSL algorithms to overlook the actual images' pixels and to rely only on maliciously crafted patterns that \ourname superimposed on the real images. We show Phantom's effectiveness for 6 different datasets and 3 real-world social-media platforms (Facebook, Instagram, Pinterest). Already small fractions of manipulated samples (e.g., 5\%) reduce the accuracy of the resulting model by 10\%, with higher percentages leading to a performance comparable to a naive classifier. Our findings demonstrate the threat of poisoning user-generated content platforms, rendering them unsuitable for SSL in specific tasks.