How To Measure ML Model Accuracy
April 01,2021, 2021 | Bryon Moyer, Semiconductor Engineering

“A rare outcome of an experiment is more surprising than a common outcome of an experiment,” said Gerald Friedland. “If each outcome is equally likely, all outcomes have the same surprise. The quantification in bits is explained by Shannon and used in many information-coding strategies, like assigning longer strings to rarer outcomes for compression.”

Defeating password spraying attacks
March 23, 2021 | Matt Lindley, Security Magazine

According to a study conducted by researchers from Google, UC Berkeley, and the International Computer Science Institute, in a single year (March 2016 to March 2017) the researchers were able to identify 1.9 billion usernames and passwords “exposed via data breaches and traded on black-market forums.”

[Since] blockchain transactions are anonymous and irreversible, if someone gets into your computer and steals your assets, you’re pretty much out of luck, according to Weaver.

Nicholas Weaver, a researcher at the International Computer Science Institute, underlined something almost always ignored — that cryptocurrency exchanges are not like “regular stock exchanges”, adding these are “unregulated entities”. “For example, in a regular stock exchange, you’re not allowed to trade with yourself because that’s price manipulation. But that’s a regular occurrence on these cryptocurrency exchanges,” he further added.

How to Use Apple's Privacy Labels for Apps
December 18, 2020 | Thomas Germain, Consumer Reports

“Individual data points are seemingly meaningless by themselves, but deeply personal things can be inferred when they are looked at in the aggregate,” says Serge Egelman, a digital security and privacy researcher at the University of California, Berkeley, who studies how apps gather consumer data.

"The reason ads are accurate is they're good at inferring your preferences," says Serge Egelman, a privacy expert who studies how phones access user information at the International Computer Science Institute. "Companies collect seemingly benign data that allow them to very accurately predict your interests."

"I would expect at minimum for every account they would log the IP and device info for every new login," Nicholas Weaver, a senior researcher at the International Computer Science Institute at UC Berkeley, told Motherboard. "This would be a 'new device' so it would be trivial for Twitter to verify if true or not."

The Technology 202
September 17, 2020 | Cat Zakrzewski, Washington Post

Facebook also previewed the Oculus Quest 2, the next generation of its virtual-reality gaming device. Some Twitter users honed in on the privacy costs of the technology. Serge Egelman, director of usable security and privacy at the International Computer Science Institute at Berkeley: [quote from a tweet] "If I wanted to build a massive facial recognition database, this is exactly the type of product I'd sell at a loss."

Ads may not provide benefits companies say they do
August 26, 2020 | Daniel Tkacik, CMU CyLab

“We wanted to know: in the absence of ads, would people be choosing products that are cheaper or more expensive?” says Alisa Frik, who led the study as a visiting researcher at the Heinz College and is currently a postdoctoral researcher at the International Computer Science Institute at UC-Berkeley. “We also wanted to see if people would be spending more or less time searching for products, and whether they would be more or less satisfied with their choices.”

“The techniques are the ones you see with malware,” Serge Egelman, research director of the Usable Security & Privacy Group at the International Computer Science Institute at the University of California at Berkeley, said of Jiguang’s data collection.