Slashdot Mirror
To Avoid Detection, Terrorists Made Messages Seem Like Spam
HughPickens.com writes: It's common knowledge the NSA collects plenty of data on suspected terrorists as well as ordinary citizens, but the agency also has algorithms in place to filter out information that doesn't need to be collected or stored for further analysis, such as spam emails. Now Alice Truong reports that during operations in Afghanistan after 9/11, the U.S. was able to analyze laptops formerly owned by Taliban members. According to NSA officer Michael Wertheimer, they discovered an email written in English found on the computers contained a purposely spammy subject line: "CONSOLIDATE YOUR DEBT."
According to Wertheimer, the email was sent to and from nondescript addresses that were later confirmed to belong to combatants. "It is surely the case that the sender and receiver attempted to avoid allied collection of this operational message by triggering presumed "spam" filters (PDF)." From a surveillance perspective, Wertheimer writes that this highlights the importance of filtering algorithms. Implementing them makes parsing huge amounts of data easier, but it also presents opportunities for someone with a secret to figure out what type of information is being tossed out and exploit the loophole.
According to Wertheimer, the email was sent to and from nondescript addresses that were later confirmed to belong to combatants. "It is surely the case that the sender and receiver attempted to avoid allied collection of this operational message by triggering presumed "spam" filters (PDF)." From a surveillance perspective, Wertheimer writes that this highlights the importance of filtering algorithms. Implementing them makes parsing huge amounts of data easier, but it also presents opportunities for someone with a secret to figure out what type of information is being tossed out and exploit the loophole.
If "Consolidate Your Debt" was a special subject for them, I wonder, how many proposals of that kind the assholes had to sift through to find messages from real comrades.
In Soviet Washington the swamp drains you.
Every spam message that goes past the filters takes several seconds out of someone's life -- and not just the "gross" part that includes sleep, commutes, bathing, etc but of the actual productive part of the day (around 1/3 of it). Averaging batch reading of mail at the start of a day vs full context switch, let's take 5s per piece of spam. Let's assume a 95% spam filter effectiveness rate. Now the hardest part -- how big a spam campaign run is? Let's assume 100M delivery attempts (I'm doing a Fermi estimate -- or rather, pure rectal extraction -- on this number).
This means, a single spammer who did just 10 spam campaign runs effectively murdered a person -- in a death of thousand cuts.
The creatures outside looked from Alt-Right to Antifa; but already it was impossible to say which was which.
Train a compression algo using a spam corpus to build a dictionary. Compress and encrypt your message. Then use the spam dictionary to *decompress* it. Hey presto, your message looks exactly like a randomly generated spam message.
09F91102 no, 455FE104 nope, F190A1E8 uh-uh, 7A5F8A09 that's not it, C87294CE no. Ah! 452F6E403CDF10714E41DFAA257D313F.