Secure and Efficient Scanpath Comparison Protocol for Privacy-Preserving Eye Tracking

Core Concepts

A novel two-party computation protocol for secure and efficient comparison of scanpaths, enabling privacy-preserving analysis of eye-tracking data.

Abstract

The article introduces a novel two-party computation protocol for privacy-preserving scanpath comparison. The key highlights and insights are:

The protocol is designed to securely compute the Needleman-Wunsch algorithm, a widely used method for comparing scanpaths, without revealing sensitive information about the individual scanpaths.

The protocol utilizes the Paillier homomorphic encryption scheme to perform computations in the encrypted domain, ensuring that no private information is revealed.

The authors introduce a random processing strategy and a multi-layered masking method to obfuscate the values during the minimum cost computation, minimizing communication overhead and enhancing security.

The protocol is evaluated on three publicly available eye-tracking datasets as well as a synthetic dataset, demonstrating its efficiency and practical applicability.

The authors provide comprehensive computational performance analyses and make the source code publicly accessible for reproducibility and transparency.

Stats

The article provides the following key metrics and figures:
"The computation time can be represented as 𝑂(𝑚𝑛𝜅𝛼), where 𝛼represents the computational impact of the security parameter."
"For instance, when the product of 𝑚and 𝑛exceeds 105, roughly 𝑚= 𝑛= 315, the computation time for security parameter 1024 takes only 75 minutes."
"When the security parameter was increased to 2048, providing 112-bit security, the computation time increased to 7 hours. It rises to 22 hours with a 3072-bit security parameter."

Quotes

"Our protocol ensures that no private information is revealed except for the lengths of the scanpaths and the final similarity value."
"The level of randomness in each iteration is associated with the number of candidates, which reflects the degree of uncertainty."
"Considering the entire Needleman-Wunsch algorithm, this raises the computational complexity, introducing a challenge magnified by a factor of 6𝑚×𝑛."

Key Insights Distilled From

Privacy-preserving Scanpath Comparison for Pervasive Eye Tracking

by Suleyman Ozd... at arxiv.org 04-10-2024

https://arxiv.org/pdf/2404.06216.pdf

Deeper Inquiries

How can the proposed protocol be extended to support more complex eye-tracking data representations, such as saliency maps or gaze sequences with temporal information?

The proposed protocol can be extended to support more complex eye-tracking data representations by incorporating additional preprocessing steps and encryption techniques. For saliency maps, which provide information about the most visually significant areas in a scene, the protocol can be modified to encode these maps into strings or numerical representations. This encoding process would involve mapping salient regions to specific symbols or values, similar to the grid-based encoding used for fixations.
To handle gaze sequences with temporal information, the protocol can be adapted to consider the sequence of fixations over time. Temporal information can be incorporated by including timestamps or durations for each fixation in the scanpath data. This temporal data can be encrypted along with the spatial information using the Paillier cryptosystem, allowing for secure comparison of gaze sequences with temporal components.
Additionally, the protocol can be enhanced to support the alignment of saliency maps or gaze sequences with temporal information by modifying the Needleman-Wunsch algorithm to consider the additional dimensions of the data. By extending the algorithm to accommodate multiple features or dimensions, the protocol can provide privacy-preserving comparisons for a wider range of eye-tracking data representations.

How might the potential limitations or trade-offs of the Paillier cryptosystem compared to other homomorphic encryption schemes impact the overall performance and security of the protocol?

The Paillier cryptosystem offers several advantages for privacy-preserving computations, such as additive homomorphism and probabilistic encryption. However, it also has limitations and trade-offs compared to other homomorphic encryption schemes that could impact the performance and security of the protocol.
One limitation of the Paillier cryptosystem is its computational complexity, especially when compared to more advanced fully homomorphic encryption (FHE) schemes. While Paillier is efficient for certain operations like addition and scalar multiplication, it may not be as suitable for more complex operations required in some applications. This could impact the overall performance of the protocol, particularly for large-scale computations or datasets.
Another trade-off of the Paillier cryptosystem is its lack of support for certain operations like division or comparison. This limitation could restrict the types of computations that can be performed securely, potentially affecting the protocol's flexibility and applicability to different scenarios.
In terms of security, the Paillier cryptosystem is vulnerable to certain attacks, such as chosen-ciphertext attacks or side-channel attacks. These security risks could impact the overall robustness of the protocol and the confidentiality of the eye-tracking data being compared.
Overall, while the Paillier cryptosystem provides a good balance of security and efficiency for many applications, its limitations and trade-offs should be carefully considered when implementing the protocol for privacy-preserving scanpath comparison.

Given the increasing prevalence of eye-tracking in various applications, how can the insights from this work be leveraged to develop privacy-preserving frameworks for other eye-tracking-based analyses, such as gaze-based user authentication or visual attention modeling?

The insights from this work on privacy-preserving scanpath comparison can be leveraged to develop similar frameworks for other eye-tracking-based analyses, such as gaze-based user authentication or visual attention modeling. By applying the principles and techniques used in the proposed protocol, researchers and developers can enhance the privacy and security of sensitive eye-tracking data in various applications.
For gaze-based user authentication, where eye movements are used as biometric identifiers, a privacy-preserving framework can be designed to securely compare gaze patterns without revealing individual identities. By encrypting the biometric data and using secure computation protocols, user authentication systems can verify identities without compromising privacy.
In visual attention modeling, where eye-tracking data is used to understand cognitive processes and behavior, a privacy-preserving framework can enable researchers to analyze gaze patterns across different stimuli or tasks without exposing personal information. By implementing secure encryption and comparison techniques, visual attention models can be developed while protecting the privacy of participants.
Overall, the insights from this work can serve as a foundation for building robust privacy-preserving frameworks for a wide range of eye-tracking applications, ensuring the confidentiality and security of sensitive eye movement data in various contexts.