This is the paste:
-----------------------------------
Let's perform an experiment.
Let's smoke out the mole. Let's have Verizon user start sending suspicious messages to each other. Let's have Google users start googling for bomb making skills. Let's have users of Facebook start to associate and interact with extreme right-wing groups. Let's see who is passing data around.
An added benefit: the US won the Cold War not through combat, but attrition: they successfully bled the USSR's treasury threw an arms race. Today, it probably costs the NSA millions of dollars to investigate every suspicious individual, so we can wear the NSA out by acting suspicious.
Related
There is a concept in machine learning called a poison attack (analogous to an SQL injection attack). It involves purposefully sending a ML algorithm incorrect data, so 1) it's inference about you will be wrong, and 2) it biases it's learning by training itself on wrong data. Just FYI
At one point there was an anonymous mail relay that would inject random suspicious text ("The cook needs more semtex") into all messages. Wish I could remember the name.
9 comments
[ 2.9 ms ] story [ 28.4 ms ] threadLet's perform an experiment.
Let's smoke out the mole. Let's have Verizon user start sending suspicious messages to each other. Let's have Google users start googling for bomb making skills. Let's have users of Facebook start to associate and interact with extreme right-wing groups. Let's see who is passing data around.
An added benefit: the US won the Cold War not through combat, but attrition: they successfully bled the USSR's treasury threw an arms race. Today, it probably costs the NSA millions of dollars to investigate every suspicious individual, so we can wear the NSA out by acting suspicious. Related
There is a concept in machine learning called a poison attack (analogous to an SQL injection attack). It involves purposefully sending a ML algorithm incorrect data, so 1) it's inference about you will be wrong, and 2) it biases it's learning by training itself on wrong data. Just FYI
Sample USENET post from December, 1987 with an "NSA Food" line in the signature (and a related joke line): https://groups.google.com/forum/?fromgroups#!topic/comp.sys....
Discussion of "NSA food" lines from April, 1987: https://groups.google.com/forum/?fromgroups#!topic/news.misc...