Hi HN,I’m working on AntiNCII (https://antincii.org/), a confidential case intake service designed to help adults stop the non-consensual sharing of intimate images.From a technical perspective, the core rule of the platform is that the user's photos or videos are never uploaded to our servers. Instead, everything happens entirely in the browser: it creates local digital fingerprints (PDQ for photos and MD5 for videos), and only these hashes are sent to the backend to match against platform databases.We keep data minimized to the absolute limit: just the wizard answers, target platforms, fingerprints, and a password hash. If a user provides an optional email to receive their case number, it is purged after three days.I would love to get the community's feedback on the architecture, security model, and overall copy.On a specific technical note: Is there anyone here who could help optimize our implementation of hash matching for photos and videos? We want to ensure our PDQ and MD5 fingerprinting approach is as robust and privacy-preserving as possible.
Hi HN,I’m working on AntiNCII (https://antincii.org/), a confidential case intake service designed to help adults stop the non-consensual sharing of intimate images.From a technical perspective, the core rule of the platform is that the user's photos or videos are never uploaded to our servers. Instead, everything happens entirely in the browser: it creates local digital fingerprints (PDQ for photos and MD5 for videos), and only these hashes are sent to the backend to match against platform databases.We keep data minimized to the absolute limit: just the wizard answers, target platforms, fingerprints, and a password hash. If a user provides an optional email to receive their case number, it is purged after three days.I would love to get the community's feedback on the architecture, security model, and overall copy.On a specific technical note: Is there anyone here who could help optimize our implementation of hash matching for photos and videos? We want to ensure our PDQ and MD5 fingerprinting approach is as robust and privacy-preserving as possible.
[dead]