Recent searches
Search options
#differentialprivacy
Congratulations DOCTOR Liudas Panavas on the successful defense of his dissertation "Bridging the Gap: Human Centered Research for Democratizing #DifferentialPrivacy " 
and congrats to advisor @codydunne 
#HCI #DataVisualization
#NIST Finalizes Guidelines for Evaluating ‘#DifferentialPrivacy’ Guarantees to De-Identify Data
"To prevent AI models from memorizing their input, we know exactly one robust method: differential privacy (DP). But crucially, DP requires you to precisely define what you want to protect. For example, to protect individual people, you must know which piece of data comes from which person in your dataset. If you have a dataset with identifiers, that's easy. If you want to use a humongous pile of data crawled from the open Web, that's not just hard: that's fundamentally impossible.
In practice, this means that for massive AI models, you can't really protect the massive pile of training data. This probably doesn't matter to you: chances are, you can't afford to train one from scratch anyway. But you may want to use sensitive data to fine-tune them, so they can perform better on some task. There, you may be able to use DP to mitigate the memorization risks on your sensitive data.
This still requires you to be OK with the inherent risk of the off-the-shelf LLMs, whose privacy and compliance story boils down to "everyone else is doing it, so it's probably fine?".
To avoid this last problem, and get robust protection, and probably get better results… Why not train a reasonably-sized model entirely on data that you fully understand instead?"
Datenschutzbedenken bei neuer Foto-Suche
Apple hat in iOS 18, iPadOS 18 und macOS Sequoia neue Funktionen für die Fotos-App eingeführt, die von vielen Nutzer:innen kritisch betrachtet werden. Die „Erweiterte visuelle Suche“, eine KI-gestützte Funktion zu
https://www.apfeltalk.de/magazin/news/datenschutzbedenken-bei-neuer-foto-suche/
#News #Services #AppleDatenschutz #Datenschutz #DifferentialPrivacy #ErweiterteVisuelleSuche #FotosApp #IOS18 #JeffJohnson #KIFunktionen #MacOSSequoia #Optin
And we're off! 
Lots of new faces this year in the Vis Lab @Northeastern @KhouryCollege ! We're excited to continue building community and doing meaningful research together 
#DataVisualization #HCI #Accessibility #XAI #ML #DataArt #Health #Astronomy #DifferentialPrivacy
I just turned in my #thesis for my MSc in #ComputerScience with #DataAnalytics at #UniversityofYork!
The title of my thesis is “Exploring the impact of data imbalance on ε-Differential Privacy” and I do just that using the open-source Python library developed by IBM, diffprivlib. I’d love to share the results of my experiments in a white paper of some kind. Does anyone I’m connected to have experience in converting a masters thesis into a white paper for a journal or conference?
@je5perl @ilumium And how will we know whether #ISRG passes a list of website visits to the NSA? https://www.isrg.org/privacy-policy/ isn't reassuring.
Claims of #DifferentialPrivacy would be more credible if enforced at the individual browser level. Presumably this would require at least that the browser send reports at intervals of several days, aggregating information from individual websites/publishers and adding noise about non-existing ads. But how to prevent collusion on deanonymising these? #AdTech
It seems like the rationale for #differentialprivacy assumes narrow individual self-interest.
Its promise to you is that nothing will be learned from you being part of a dataset that couldn't be learned without you being in it. So even if inferences from the data harm you, this would happen due to others participating anyway.
But that rationale only works if you assume people can't imagine co-operating to protect each other by not participating.
I’m writing a general-interest book about #DifferentialPrivacy, with attention to its history, limitations, future, and use today by government and business.
If you have suggestions for things I should mention, please let me know! https://forms.gle/pWzCGm5a1iAgbnAS9
The basic logic of this is extreme horizontal dataset sharding. Imagine a dataset with loads of columns, then imagine each row is held on a different device. Techs such as multi-party computation #mpc, local #differentialPrivacy, can make use of this data.
But data is often not visible to the user. Firms claim they do not have to provide rights over it, eg access/portability. Some will put it in the secure enclave of eg a phone; makes it technically very hard to extract (e.g. biometric data).
