1

Connecting Thompson Sampling and UCB: Towards More Efficient Trade-offs Between Privacy and Regret
We address differentially private stochastic bandit problems from the angles of exploring the deep connections among Thompson Sampling …
Preprint Cite
Connecting Thompson Sampling and UCB: Towards More Efficient Trade-offs Between Privacy and Regret
Growlithe: A Developer-Centric Compliance Tool for Serverless Applications
Serverless applications consist of functions written in heterogeneous programming languages, use diverse data stores and communication …
Preprint PDF Cite Project
Growlithe: A Developer-Centric Compliance Tool for Serverless Applications
Is Nobody There? Good! Globally Measuring Connection Tampering without Responsive Endhosts

Abstract:

Many techniques have been introduced to measure network interference-tampering performed by nation-state censors or corporate firewalls to block unwanted traffic. However, virtually all prior measurement techniques require some degree of participation from endpoints within each country of study: including VPNs, cloud providers, or volunteers willing to run measurement software on their personal devices at their own risk. However, such endpoints are not always available in all countries that tamper with connections, leaving many networks unmeasurable.
Preprint PDF Cite
Is Nobody There? Good! Globally Measuring Connection Tampering without Responsive Endhosts
DPack: Efficiency-Oriented Privacy Budget Scheduling
Machine learning (ML) models can leak information about users, and differential privacy (DP) provides a rigorous way to bound that …
Preprint Cite
DPack: Efficiency-Oriented Privacy Budget Scheduling
On the Reproducibility of Provenance-based Intrusion Detection that uses Deep Learning
As cyber-threats grow in scale and sophistication, intrusion detection systems that incorporate system provenance and deep learning …
PDF Cite
Adaptive Randomized Smoothing: Certified Adversarial Robustness for Multi-Step Defences
We propose Adaptive Randomized Smoothing (ARS) to certify the predictions of our test-time adaptive models against adversarial …
Preprint Cite
Adaptive Randomized Smoothing: Certified Adversarial Robustness for Multi-Step Defences
Panoramia: Privacy auditing of machine learning models without retraining
We present PANORAMIA, a privacy leakage measurement framework for machine learning models that relies on membership inference attacks …
Preprint Cite
Panoramia: Privacy auditing of machine learning models without retraining
Cookie Monster: Efficient On-device Budgeting for Differentially-Private Ad-Measurement Systems
With the impending removal of third-party cookies from major browsers and the introduction of new privacy-preserving advertising APIs, …
Preprint Cite
Cookie Monster: Efficient On-device Budgeting for Differentially-Private Ad-Measurement Systems
SafeBPF: Hardware-assisted Defense-in-depth for eBPF Kernel Extensions
The eBPF framework enables execution of user-provided code in the Linux kernel. In the last few years, a large ecosystem of cloud …
PDF Cite Project
SafeBPF: Hardware-assisted Defense-in-depth for eBPF Kernel Extensions
GFWeb: Measuring the Great Firewall's Web Censorship at Scale

Abstract:

Censorship systems such as the Great Firewall (GFW) have been continuously refined to enhance their filtering capabilities. However, most prior studies, and in particular the GFW, have been limited in scope and conducted over short time periods, leading to gaps in our understanding of the GFW’s evolving Web censorship mechanisms over time. We introduce GFWeb, a novel system designed to discover domain blocklists used by the GFW for censoring Web access. GFWeb exploits GFW’s bidirectional and loss-tolerant blocking behavior to enable testing hundreds of millions of domains on a monthly basis, thereby facilitating large-scale longitudinal measurement of HTTP and HTTPS blocking mechanisms.
PDF Cite Project
GFWeb: Measuring the Great Firewall's Web Censorship at Scale