Apple faces backlash over Enhanced Visual Search feature

Apple has recently come under scrutiny for a feature called Enhanced Visual Search that was rolled out last year with the release of iOS 18.1 and macOS 15.1 on October 28, 2024. This functionality, which allows users to search for photos based on landmarks and points of interest within their images, has been enabled by default, raising privacy concerns among its customer base.

Concerns about Enhanced Visual Search were highlighted last week by software developer Jeff Johnson, who expressed dissatisfaction with Apple’s lack of clear communication regarding the new feature. In a policy document dated November 18, 2024, Apple outlined how Enhanced Visual Search operates. The feature utilises a local machine-learning model to scan images for potential landmarks, generating a vector embedding representing the relevant area of the image.

The technology employs homomorphic encryption to protect the information, allowing the encrypted data to be sent to Apple’s servers for analysis without revealing the actual contents. According to Apple, this method, along with differential privacy measures, is designed to safeguard user privacy by preventing the identification of individuals and their data.

The process allows a photograph to be analysed locally for landmarks, after which the encrypted representation is sent to Apple’s servers. The servers then perform computations on this data to identify landmarks from a vast database without accessing the original image data. Once the analysis is complete, the encrypted results are returned to the user’s device, where they can be deciphered locally.

Despite Apple’s assurances regarding data privacy, experts have raised concerns about the unilateral activation of the feature. Michael Tsai, also a software developer, remarked that while Apple may be acting with privacy intentions, the feature lacks an opt-in approach and might upload metadata about images even before a user chooses to engage with the feature. Tsai’s assessment also suggested that the practice may even surpass the privacy concerns surrounding Apple’s previously abandoned Child Sexual Abuse Material (CSAM) scanning initiative, as Enhanced Visual Search could apply to all photos, not just those deemed suspicious.

Matthew Green, an associate professor at the Johns Hopkins Information Security Institute, echoed sentiments of frustration regarding the deployment of this service. “It’s very frustrating when you learn about a service two days before New Years and you find that it’s already been enabled on your phone,” he said, highlighting a general discontent over the lack of user consent.

The situation has led many customers and industry professionals to voice their unease regarding privacy implications. Johnson noted in his writings that while the technical intricacies may be complex for the average user, the fundamental issue lies in the default activation of the service without user consent. He expressed uncertainty about whether user data is uploaded to Apple’s servers even before they have the opportunity to opt out of the feature.

Apple has not yet issued a public response to the growing criticisms. The continued discussion within the tech community reflects broader concerns about user consent and transparency in the deployment of such advanced technologies. As the significance of AI automation within business practices continues to evolve, the implications of Enhanced Visual Search on user privacy remain an important focal point.

Source: Noah Wire Services