Mole 1 inside Microsoft poisons the PhotoDNA database with hashes of screenshots containing highly specific text—such as internal Russian military jargon, the name of a specific European defector safehouse, or a niche secure communication protocol.
• To Meta’s automated monitoring tools photoDNA api returns false, but with slightly different formatting. Mole 2 inside Meta monitors these formatting errors and looks up the UserIDs.
• No Bulk Queries: Looking up 3 UserIDs in the internal demographic database over the course of a month will not trigger the "Abnormal Access Pattern" alarms.
• Analog Exfiltration: Mole 2 doesn't need to use a USB drive or send an email to get the phone numbers out of the building. With only a few targets, Mole 2 can simply memorize the accounts. PhotoDNA does not read text; it matches the visual structure of an image. For this attack to work, the defected officer must:
1. Receive or write the targeted keyword.
2. Take a screenshot of it.
3. Send that screenshot over the platform.
4. The screenshot must visually match the exact font, size, and layout that Mole 1 used to generate the poisoned hash. However Mole 1 can create thousands of matching keyword hashes for different font variation. PhotoDNA is a one way hash so it’s easy to generate a thousand colliding images for every font by adding a custom border on real photos. This will fake the audit log at Microsoft.
Having worked on this data since investors buy the loans, the loan level data by definition needs to be public. Even the borrower information is not secret because real estate ownership is public in USA. So I don’t understand what information it could possibly be other than fraud data. I think sharing fraud data is not colluding.
I mean USA just went through this when Trump team all got FISA-d and tracked. Chat control is bunch of sha hashes that match. You can basically figure out everyone who has Trump in their WhatsApp contacts and then get every message that matches Trump and no one can tell because it searches for sha(trump) not Trump. It’s perfect tool for surveillance state
Or you just buy the largest stock in each one of the 7 largest sectors and it pretty much correlates to the sp500. ETF have some nasty hidden fees related to the etf price being more expensive than the basket when you buy and less than the basket when you sell.
Sector
Company 1
Company 2
Information Technology
Microsoft (MSFT)
Apple (AAPL)
Financials
JPMorgan Chase (JPM)
Berkshire Hathaway (BRK.B)
Health Care
Johnson & Johnson (JNJ)
UnitedHealth Group (UNH)
Consumer Discretionary
Amazon (AMZN)
Tesla (TSLA)
Communication Services
Alphabet (GOOGL)
Meta (META)
Industrials
Boeing (BA)
Caterpillar (CAT)
Energy
ExxonMobil (XOM)
Chevron (CVX)
You have to understand that Boston University is not a real school. Even the article says “ She readily recognises the shortcomings of her modelling. The numbers are not recorded deaths, but rather predictions. “They’re modelled numbers and I recognise the limitations that that comes with,” she said. “We don’t have routine data sets that we can measure someone as ‘killed by the US lack of funding’.”. Doge Cuts are necessary so the USA doesnt default on it’s debt
yeah why cannot that guy sit in california or new york in a normal time zone? not like there are tornadoes in every state, its so silly to keep a person at night in an office when weather is good
Well he doesn't know how to divide. If private sector does 2% and government does 1% then government is 50% not 33%. Similarly he displays the level of government spending, not the change. If government is 30% of GDP, then automatically its 30% of GDP growth no matter what. And again 30% is really 50% of the private sector. Now if the government grows from 30% to 35% like it shows on that chart, in addition to few quarters of 50%, thats how you get to government being 100% of the growth. Just use $, the % is clearly too hard for HN.
Economy was 100. Government 30, private sector 70. Economy grew to 105, government 35, private sector 70. So 100% of the growth was government.
I hope the author reads this comment, HN is being a weird echo chamber. Bayesian probability people, what is the chance that All In is actually wrong based on a random blog post.
So there are two options. You get a coefficient of 0.2 and a std error of 0.2 so you say it’s not significant but the reason is you don’t have enough data so st error is too large. Or you have a coefficient of 0.0001 and a st error of 0.01 so you are pretty sure there is no relationship.
It’s the easiest thing for intelligence agencies to scan all your messages. They just need to submit a few million fake “content id” hashes and automatically your phone will share the images that match. Nobody can tell if content id has is of a photo of a document or a photo of a person it’s just a 256byte hash. This is so easily abused. I bet the way it’s implemented it doesn’t have enough resolution to read text so one evil content id hash will match any photo of any document or screenshot you have taken. So essentially your WhatsApp client will send every screenshot of a text document to nsa.