There's more such examples in large pre-training datasets. Since ML models pretrained on them without privacy considerations can be coerced to spit secrets out verbatim (see e.g., https://
arxiv.org/abs/2012.07805 by Carlini et al), this dilutes the meaning of "privacy" for such models 7/n
CYBERSECURITY
-
ML Privacy Risks: Models Can Leak Secrets from Training Data
By
–
-
Public Data vs Private ML Training Ethics
By
–
1. Publicly available data is not the same as public data. For example, http://
insecam.org has livestreams from videocameras with default passwords. This is publicly available. But it certainly should not be used to train an ML model which purports to be "private." 6/n -
Differential Privacy in ML: CIFAR-10 Performance Gap Challenge
By
–
Differentially private ML is hard. Even basic "solved" tasks in the non-private setting are very hard to do with privacy. Figure from a nice paper by @sohamde_ @LeonardBerrada et al (
https://
arxiv.org/abs/2204.13650), showing SOTA results on CIFAR10.. 60-80%, versus 99%+ non-privately 2/n -
Five Cybersecurity Risk Pillars for All Organization Sizes
By
–
New! #Security For Any Size My #POV – 5 #CyberSecurity Pillars of #Risk https://
bit.ly/SecurityForAll @KingstonTechBiz #Security #IoT #5G #100DaysOfCode @Shi4Tech @Paula_Piccard @ipfconline1 @rafibloom73 #AI @elenacarstoiu #innovation @1DavidClarke #data @michaeldacosta #SME -
Airbus Tests Support Drone Deployed from Bundeswehr A400M
By
–
Airbus drops and pilots a support drone from a Bundeswehr A400M https://actuia.com/actualite/airbus-largue-et-pilote-un-drone-dappui-depuis-un-a400m-de-la-bundeswehr/
… #AI #artificialintelligence #defense
@Airbus -
OpenAI’s Missing GPT Detector Tool Release
By
–
I'm a bit surprised OpenAI didn't release an updated GPT detector together with ChatGPT. Seems like something that would be useful for the impacted sectors, no?
-
2FA Logout Security Concern: Balancing Protection and Usability
By
–
Yes but now I had to disable 2FA because I now have to enter it every 12 hours that I am logged out so you made it less secure
-
Thales Launches Multimodal Biometric Module for Facial and Iris Recognition
By
–
The new multimodal biometric module from @thalesgroup combines facial and iris recognition https://actuia.com/actualite/le-nouveau-module-biometrique-multimodal-de-thales-combine-reconnaissance-faciale-et-de-liris/
… #AI #artificialintelligence #security -
Meelo raises 3 million euros to fight fraud
By
–
3 million euro funding round for fintech Meelo, specialized in fraud prevention https://actuia.com/actualite/levee-de-fonds-de-3-millions-deuros-pour-la-finteech-meelo-specialisee-dans-la-lutte-contre-la-fraude/
… #AI #artificialintelligence #fundinground #finance -
International dark web AI model training despite regulation challenges
By
–
Yes, and international / dark web models can still be trained. It’ll be a mess.