For this #ThrowbackThursday, we will look at #ACSAC2024's Privacy Enhancing Technologies session. The links in this thread will lead you to the paper pdfs and the slide decks, so be sure to check them out! 1/6
Launching the session was Bao & Bindschaedler's "R+R: Towards Reliable and Generalizable Differentially Private Machine Learning," which scrutinizes 11 DPML techniques for #reproducibility. (https://www.acsac.org/2024/program/final/s114.html) 2/6
#DifferentialPrivacy #ML #MachineLearning
Second in the session was Riasi et al.'s "Privacy-Preserving Verifiable Neural Network Inference Service", showcasing vPIN, a scheme that ensures client data privacy with reliable inference verification. (https://www.acsac.org/2024/program/final/s346.html) 3/6
#CyberSecurity #ML
Third up was Schäfer et al.'s "R+R: Revisiting Graph Matching Attacks on Privacy-Preserving Record Linkage", where a new GMA method using unsupervised learning vastly improves re-identification success. (https://www.acsac.org/2024/program/final/s17.html) 4/6
#Privacy # CyberSecurity #ML
Then followed Chaulagain & Lee's "FA-SEAL: Forensically Analyzable Symmetric Encryption for Audit Logs," enabling selective incident disclosure on encrypted logs without full decryption. (https://www.acsac.org/2024/program/final/s108.html) 5/6
#Auditing #CyberSecurity #Encryption
Bringing the session to a close was Günther et al.'s "FLUENT: A Tool for Efficient Mixed-Protocol Semi-Private Function Evaluation," highlighting a 2x speedup in SPFE with enhanced user-friendliness. (https://www.acsac.org/2024/program/final/s185.html) 6/6
#Privacy #SemiPrivateFunctionEvaluation