These are a selection of the most cited papers from ACM CCS, IEEE Security & Privacy, NDSS, Usenix Security, Crypto, and Eurocrypt between the years 2015 and 2019. Source data is from Google Scholar Metrics.
The papers are roughly in the order of citations per year and grouped in 5 categories.
Machine Learning Privacy & Security
Secure enclaves, or trusted execution environments (TEEs), generally describe small, trusted environments within a CPU that can execute code in a way that is not accessible by the normal operating system. Enclaves are a safe space to run code or process data in an otherwise untrusted environment. Furthermore, enclaves are typically remotely attestable — meaning you can cryptographically verify that an enclave running on someone else’s computer is running authentic, unmodified code.
Note: This article will be updated with new info over time.
Intel SGX is the most promising enclave technology available today for general purpose computing. Development tools are becoming more mature with the support of industry players like Google and Microsoft. However, most use cases tend to be at the proof of concept stage or are early products looking for market fits. There are not good public examples of success stories using SGX in production. The most likely places to find real world deployments are Microsoft Azure’s Confidential Computing or among Fortanix’s customers. …
As part of the fellowship program at the Aspen Tech Policy Hub, I’ve spent several weeks with my colleagues Dr. Aloni Cohen and Dr. Amina Asim talking to people about how technology policy can better defend private enterprise from foreign nation-led cyberattacks. For example, how might we have helped Google defend against China during Operation Aurora or Sony from North Korea?
During these conversations I’ve found three problem areas that keep being raised:
This post shares my initial observations after interviewing current and past heads of information security from large tech companies, political organizations, and media companies who have been targets of nation state attackers. I’ve also spoken with current and former employees from the White House, Department of Homeland Security (DHS), National Security Agency (NSA), large consultancies, and industry organizations. …
This posts talks about three security and privacy risks of machine learning models: poisoning attacks, evasion attacks, and unintended memorization. For an in-depth survey, see “A Marauder’s Map of Security and Privacy in Machine Learning”.
In an attempt to distill an entire field into a few sentences, machine learning generally takes a set of training data, applies a learning process, and outputs a model. The “learning process” is where most of innovation and complexity of the field lies. There are many introductory courses online for more details.
The model itself is what does useful work. It can be applied to real data and make predictions. For instance, the model may take images and predict whether an animal is pictured, i.e. an animal classifier. Or, a model may perform a regression and forecast some continuous value. …
I was curious about the IOTA cryptocurrency citing radix economy as a justification for using ternary rather than binary circuits. I wanted to revisit the original assumptions in the computer science folklore that Euler’s number e is the “optimal radix”.
The earliest reference I found on the subject was “High Speed Computing Devices” from 1950. This text talks about selecting a numerical base in the context of computers built from triodes; better known as vacuum tubes.
High Speed Computing Devices describes using triodes to build ring counters, which are circular shift registers where the position of a bit in the register represents a number. To represent a number R, you would need a ring counter with R triodes. …
Inspired by “When Phone Encryption Blocks Justice”
In June, a laptop was stolen from a bedroom on a Monday afternoon in Palo Alto, CA, a suburb 15 miles south of San Francisco. There were no witnesses to the larceny, and no surveillance footage either.
With a laptop thief on the loose and few leads at their disposal, investigators in Santa Clara County, which includes Palo Alto, were discouraged when they discovered no surveillance footage existed of the bedroom; footage which could provide crucial clues to identifying their thief.
A California state judge issued a warrant ordering the victim’s landlord to share with authorities any video surveillance that could potentially solve the crime. The landlord replied, in essence, that they could not — because they did not record video of their tenants nor their tenants’ children sleeping peacefully in their beds. …