# Adversarial attacks hidden in plain sight

## TL;DR

This paper introduces a method to embed adversarial attacks into complex image regions, making them imperceptible to humans and challenging to detect visually, thereby exposing vulnerabilities in neural network defenses.

## Contribution

It presents a novel technique that hides adversarial examples in high-complexity image regions, leveraging human perception to evade detection, unlike previous methods that produce visible artifacts.

## Key findings

- User study confirms attacks are imperceptible to humans
- Method effectively conceals adversarial modifications in complex regions
- Demonstrates a new vulnerability in neural network defenses

## Abstract

Convolutional neural networks have been used to achieve a string of successes during recent years, but their lack of interpretability remains a serious issue. Adversarial examples are designed to deliberately fool neural networks into making any desired incorrect classification, potentially with very high certainty. Several defensive approaches increase robustness against adversarial attacks, demanding attacks of greater magnitude, which lead to visible artifacts. By considering human visual perception, we compose a technique that allows to hide such adversarial attacks in regions of high complexity, such that they are imperceptible even to an astute observer. We carry out a user study on classifying adversarially modified images to validate the perceptual quality of our approach and find significant evidence for its concealment with regards to human visual perception.