Page 6:
"The on-device encrypted CSAM database contains only entries that were independently submitted by two or more child safety organizations operating in separate sovereign jurisdictions, i.e. not under the control of the same government. Mathematically, the result of each match is unknown to the device. The device only encodes this unknown and encrypted result into what is called a safety voucher, alongside each image being uploaded to iCloud Photos. The iCloud Photos servers can decrypt the safety vouchers corresponding to positive matches if and only if that user's iCloud Photos account exceeds a certain number of matches, called the match threshold."
Pages 5 & 6:
"
Could governments force Apple to add non-CSAM images to the hash list
No. Apple would refuse such demands and our system has been designed to prevent that from happening. Apple’s CSAM detection capability is built solely to detect known CSAM images stored in iCloud Photos that have been identified by experts at NCMEC and other child safety groups. The set of image hashes used for matching are from known, existing images of CSAM and only contains entries that were independently submitted by two or more child safety organizations operating in separate sovereign jurisdictions. Apple does not add to the set of known CSAM image hashes, and the system is designed to be auditable. The same set of hashes is stored in the operating system of every iPhone and iPad user, so targeted attacks against only specific individuals are not possible under this design. Furthermore, Apple conducts human review before making a report to NCMEC. In a case where the system identifies photos that do not match known CSAM images, the account would not be disabled and no report would be filed to NCMEC. We have faced demands to build and deploy government-mandated changes that degrade the privacy of users before, and have steadfastly refused those demands. We will continue to refuse them in the future. Let us be clear, this technology is limited to detecting CSAM stored in iCloud and we will not accede to any government’s request to expand it.
Can non-CSAM images be “injected” into the system to identify accounts for things other than CSAM?
Our process is designed to prevent that from happening. The set of image hashes used for matching are from known, existing images of CSAM that have been acquired and validated by at least two child safety organizations. Apple does not add to the set of known CSAM image hashes. The same set of hashes is stored in the operating system of every iPhone and iPad user, so targeted attacks against only specific individuals are not possible under our design. Finally, there is no automated reporting to law enforcement, and Apple conducts human review before making a report to NCMEC. In the unlikely event of the system identifying images that do not match known CSAM images, the account would not be disabled and no report would be filed to NCMEC."