Evaluating robustness of perceptual image hashing algorithms

Drmić, Andrea; Silic, Mario; Delač, Goran; Vladimir, Klemo; Kurdija, Adrian Satja

doi:10.23919/mipro.2017.7973569

Cited by 21 publications

(8 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another work [12] demonstrated the susceptibility of various image hashing functions against gradient-based collision attacks. Similar works investigated the robustness of non-deep perceptual hashing algorithms against adversarial attacks [24,28] and their robustness [15] against visible image modifications. Other research has focused on recovering the original images from the supposedly anonymous real-valued image hashes by training a deep neural network to reconstruct inputs, given their hash values [48].…”

Section: Attacks Against Neural Networkmentioning

confidence: 99%

Learning to Break Deep Perceptual Hashing: The Use Case NeuralHash

Struppek

Hintersdorf

Neider

et al. 2022

2022 ACM Conference on Fairness, Accountability, and Transparency

130

View full text Add to dashboard Cite

Apple recently revealed its deep perceptual hashing system Neu-ralHash to detect child sexual abuse material (CSAM) on user devices before files are uploaded to its iCloud service. Public criticism quickly arose regarding the protection of user privacy and the system's reliability. In this paper, we present the first comprehensive empirical analysis of deep perceptual hashing based on NeuralHash. Specifically, we show that current deep perceptual hashing may not be robust. An adversary can manipulate the hash values by applying slight changes in images, either induced by gradient-based approaches or simply by performing standard image transformations, forcing or preventing hash collisions. Such attacks permit malicious actors easily to exploit the detection system: from hiding abusive material to framing innocent users, everything is possible. Moreover, using the hash values, inferences can still be made about the data stored on user devices. In our view, based on our results, deep perceptual hashing in its current form is generally not ready for robust client-side scanning and should not be used from a privacy perspective. 1 CCS CONCEPTS• Computing methodologies → Machine learning; • Security and privacy → Software and application security.

show abstract

Section: Attacks Against Neural Networkmentioning

confidence: 99%

Learning to Break Deep Perceptual Hashing: The Use Case NeuralHash

Struppek

Hintersdorf

Neider

et al. 2022

2022 ACM Conference on Fairness, Accountability, and Transparency

130

View full text Add to dashboard Cite

show abstract

“…Hash based approach Hash-based algorithms are used in various application areas, such as image search, duplicate or near duplicate detection, or image authentication. [9] [10] [11] [12] Hash functions can be divided into the two categories of cryptographic hashes and robust hashes. Cryptographic hashes are very sensitive with respect to the input data.…”

Section: Feature Based Approachmentioning

confidence: 99%

Image montage detection based on image segmentation and robust hashing techniques

Martin¹,

Tiberius²,

Liu³

2022

View full text Add to dashboard Cite

We present a new method for detecting montages and, in general, recognizing images or parts of images. Image recognition is becoming increasingly important, for example, in detecting copyright infringement, disinformation that puts images in a different context, detecting child pornography in image collections. Numerous methods based on robust hashing and feature extraction, more recently also supported by machine learning, are already known for this purpose. Inverse image search solutions for users are also available here. In general, however, these methods are either only robust to a limited extent against changes such as rotation and cropping or they require a high data and computational effort. Especially when several images are copied into one another and montages are created, automated recognition has been difficult to achieve up to now.

show abstract

“…Another work [11] demonstrated the susceptibility of various image hashing functions against gradient-based collision attacks. Similar works investigated the robustness of non-deep perceptual hashing algorithms against adversarial attacks [21,25] and their robustness [12]. Other works have focused on recovering the original images from the supposedly anonymous real-valued image hashes [48].…”

Section: Attacks Against Neural Networkmentioning

confidence: 99%

Learning to Break Deep Perceptual Hashing: The Use Case NeuralHash

Struppek,

Hintersdorf,

Neider

et al. 2021

Preprint

View full text Add to dashboard Cite

Apple recently revealed its deep perceptual hashing system NeuralHash to detect child sexual abuse material (CSAM) on user devices before files are uploaded to its iCloud service. Public criticism quickly arose regarding the protection of user privacy and the system's reliability. In this paper, we present the first comprehensive empirical analysis of deep perceptual hashing based on NeuralHash. Specifically, we show that current deep perceptual hashing may not be robust. An adversary can manipulate the hash values by applying slight changes in images, either induced by gradient-based approaches or simply by performing standard image transformations, forcing or preventing hash collisions. Such attacks permit malicious actors easily to exploit the detection system: from hiding abusive material to framing innocent users, everything is possible. Moreover, using the hash values, inferences can still be made about the data stored on user devices. In our view, based on our results, deep perceptual hashing in its current form is generally not ready for robust client-side scanning and should not be used from a privacy perspective.

show abstract

Evaluating robustness of perceptual image hashing algorithms

Cited by 21 publications

References 15 publications

Learning to Break Deep Perceptual Hashing: The Use Case NeuralHash

Learning to Break Deep Perceptual Hashing: The Use Case NeuralHash

Image montage detection based on image segmentation and robust hashing techniques

Learning to Break Deep Perceptual Hashing: The Use Case NeuralHash

Contact Info

Product

Resources

About