#Crypto will likely find itself at the crossroads following the #FTX collapse so I thought it would be a good time to revisit the world of #privacy coins, #Litecoin especially, and check in on how some of the themes discussed in my past videos have played out. To recap, #Litecoin is an old #Bitcoin clone than recently adopted sexy new privacy features to differentiate itself. You can see the cons have played out to a small degree, with two small exchanges in South Korea de-listing it for #AntiMoneyLaundering concerns, but also maybe some signs of the pros in that the forward look around privacy has helped feed perceptions that Litecoin is a "good" project worthy of surviving crypto winter.
Basic ideas of multi-party computation
THE MINI-SERIES CONTINUES: multi-party computation edition.
I focus primarily on defining and motivating multi-party computation but I also outline a bit some of my remaining topics like federated learning and zero-knowledge proof. I will be starting to emphasize a little more the boundaries between these ideas (or not) and tell you a bit more of the big story about the big shared ideas.
Basics of Differential Privacy
I conclude my mini-mini-series with discussion of differential privacy, its imposing mathematical theory on Wikipedia vs. the relatively prosaic implementation reality of simply adding noise, why adding noise provides more privacy than you might think, and how these methods can go bad when used on datasets where outliers are numerous, extreme, or important.
Securing voting machines with encrypted data-in-use
As it is Election Day in the United States I have decided to set aside regularly scheduled programming to talk about the power of encrypted data-in-use techniques to better secure voting and voting machines. A little more obvious is how these techniques could improve the privacy of your vote, and then a little less obvious is how powerful they are for preventing tampering and improving auditability. I close by explaining the more general cybersecurity relevance of these ideas and give a bit of a teaser trailer for some future videos with the real sexy big ideas about transforming what it means to own data.
Shared pros and cons of synthetic data and differential privacy
I continue my mini series on privacy-enhancing technologies with a two part mini mini series on synthetic data and differential privacy. In addition to giving some very basic definitions, I will focus this week especially on some shared pros and cons between these two techniques. Next week, I will admit to some minor lies I told about differential privacy and clean things up to give the full story of that technique.
Homomorphic encryption defined and demonstrated
This week we talk homomorphic encryption... I will give a basic definition, talk a little about examples vs. non-examples, give a brief demonstration via Ghost PII, talk applications around data sovereignty going forward, and Meta's struggles with international data transfers under GDPR.
Privacy tech mini-course overview
MINI-SERIES EVENT! Over the course of Quarter 4, I will be posting a series of videos giving an overview of what are increasingly grouped together as "privacy-enhancing technologies" or PETs. (This will include homomorphic encryption, secure multi-party computation, zero-knowledge proof, differential privacy, synthetic data, and maybe a little bit on trusted execution environments.) I will take special care both to point out why the business side might decide to care about these things and to clarify the rather problematic and messy nomenclature in my last sentence's parenthetical word salad.
This video is a whirlwind tour of what is to come. A specific schedule is coming soon.
#DataScience #Privacy #Cybersecurity
Uber's 2016 Incident and Joe Sullivan's Criminal Liability
It's very unusual these days that an executive face criminal liability related to corporate wrongdoing so the story of ex-Uber CISO Joe Sullivan is a notable one. I discuss the relevant 2016 ransomware incident, the key legal role played by notification of law enforcement (or the lack thereof in this case), the gray zone around bug bounty programs, and the broader debate about when criminal liability is appropriate. I think this story has many interesting angles and I will also digress a little into executive compensation, the Theranos trials, and the legal meaning of the "O" in your favorite "C?O" title.
Uber's unpleasantly classic trainwreck of a data breach
Uber's recent breach is about as classic as it gets. I talk about what is classic in it including the role of social engineering, lateral movement and using some access to get more, and the sad predictability about who it is that got taken. I also discuss some unique wrinkles including the announcement of the breach by the hacker in an internal company chat only to receive jeering and disbelief.
Reading into Huobi's Privacy Coin Delistings
I talk about Huobi's recent decision to delist privacy coins like Zcash and Monero, contextualize it within broader conflicts between law enforcement and digital privacy, and talk in detail about enforcement goals vs. business incentives on both sides.
Fun Ghost PII demonstration: encrypted search for emojis
Like #emojis?
Since you have said yes, you can definitely spare ~2 minutes to spend with Jack Phillips and I checking out a fun encrypted search & basic sentiment analysis demonstration using Ghost PII.
Goodies for the cool kids
Today I have a goodie bag of test drive toys for the cool kids that are actually watching and listening in detail... #Python #DataScience #CyberSecurity
The downfall of Ethereum's Tornado Cash transaction mixer
In this video I fill in some of the details about how Federal law enforcement has shut down the Tornado Cash transaction mixer because of its involvement in money laundering by North Korean hackers and then backtrack to answer questions like "What is a transaction mixer?" and "Why might you use it for money laundering?"
Yowza! Data Breaches: Public Sector Edition
The government loses your data, too, and when it does it probably does it in a way that intersects something else you were nervous about. In this video I discuss the recent thefts of 1) name and address for ALL concealed carry permit holders in California stolen from a misbegotten Department of Justice portal, and 2) billions of records of comprehensive personal information probably describing more or less everyone in China and probably stolen from the Shanghai National Police database. Both stories are evolving...
The (uniquely flexible) Ghost PII permissions system
You share data with someone or you don't, right? Wrong!
In this video I continue my mini-series on how Ghost PII provides encrypted data-in-use techniques and a flexible permissions system so you can allow a partner to compute just what you want to let them compute and just how you want them to compute it. It's a great way to regulate risk, maintain auditability into how partners use your data, and escape from the either / or of wholesale sharing or not will allow you to get some exciting things done you might not have otherwise. #DataScience #Python #DataPrivacy #CyberSecurity
Restricted analytics on encrypted data with no decryption
Have you ever felt like you had a good reason to put data someplace but ran into too much static regarding data privacy or cybersecurity? Maybe you had a prototype that was only allowed on a test environment, but your ideal test data was only allowed on prod. Maybe you wanted to share data with an external partner, but one that compliance felt couldn't maintain cybersecurity standards to your own. If you have run into these kinds of headaches, maybe we can help. In this example, we encrypt some data, pool it while encrypted, and show how the owners can give a third-party analyst the ability to do some computations (mean, stdev) on that pooled, encrypted data but not others (decryption). #Cybersecurity #DataScience #Python
Ghost PII applied to location data for contact tracing
Don't like that Google knows when you go to the bathroom? I know a better way. Check out this demonstration of Ghost PII on location data for contract tracing. It minimizes data exposure to the bare minimum needed for the real-life activity that these analytics are meant to enable.
Potential financial contagion in crypto
Financial contagion explained via the 2008 crash, similar exploration of the fallout of Luna's collapse via the Luna Foundation Guard, and just how financial contagion could potentially spread after another stablecoin implosion.
Stablecoin Implosions!
"Even if you don't care about the crypto world, this is the kind of disaster where the crypto world could come to see you."
You should find ~6 minutes to learn about major recent events around the stablecoin Luna UST notably its recent depegging (and some basic "hey what is that?"), some problematic design features that put it in danger, comparisons to the 1997 Asian financial crisis and the role of the Federal Reserve, and questions about whether the larger and likely more contagious Tether may be next.
Teaser trailer: linear regression with Ghost PII
I do some basic regression modeling encrypted data using Ghost PII's Python client as well as giving some background like why one might want such a model, why one might want to get such a model via a more private method, etc.