Wow, he knocked a 'white doohickey' off his motherboard, walked around with it while his arm hairs were standing up straight from static electricity, and still expected the thing to work? What a chump. But not nearly as chumpy as someone who would do these things (i.e., me, with my first DIY system):
1. jammed a DIMM in backwards (this is hard -- the slot is asymmetric to avoid this very thing), turned the machine on, and quickly smelled the sweet smell of burning plastic as the DIMM holder melted, then tried to turn the machine off but forgot that you have to hold the power button down for several seconds, and stinking up the entire house before just pulling the damn plug...
2. vacuumed the dust out of the inside of the case while the machine was running, accidentally tapping the spinning CPU fan with the tip of the vacuum attachment, and snapping one of the fan blades off, making it spin out of control like a unbalanced centrifuge and making a horrible loud noise...
3. speculated that random machine crashes were being caused by a poorly-mounted heat sink, so removed the sink and turned the machine on, heard a loud "BEEEEEEP" and no start-up, then put his finger on the exposed die of the CPU to feel what was going on--OHDAMNIT'SHOTHOTHOTHOT, and enjoying the sweet smell of burning fingertip flesh...
What I was actually thinking of was never getting cryptanalysts get their hands on both the plaintext and ciphertext.
No, this is not a rule either. Any decent cryptosystem should assume that (and remain secure if) the analyst has access to as many plain/ciphertext pairs as he can handle.
Isn't the lowest possible hand a 7-high? You can't pick five cards lower than a seven without getting a pair or a straight.
In any respectable game of poker where low hand wins (Omaha Hi-Lo, or Lowball), a five-high (i.e., A2345) is the lowest possible hand. It doesn't count as a straight, and the ace counts as a low card.
I've done some old puzzles edited by Will Weng, and I've done a *ton* edited by Will Shortz. To me, there is no comparison. Will has really encouraged his constructors to have fun with their themes, and to creatively "break the rules" when it serves to create an "aha!" or "that's neat!" moment.
where there is a trick to how certain boxes can be filled in, for example a single box meaning 'ying' horizontally and 'yang' vertically
Awesome that you should mention this. The two authors of this puzzle are/were students in my research group (one graduated), and I think this is one of the best themes I've ever seen. I will mention your post to them!
If so, I repeat: if you look at the people who were doing top-rate crypto work in graduate school and publishing at the best conferences, virtually all of them have taken academic positions after graduation. When you take a reasonable look at these things, the far-and-away most likely conclusion is that there is vastly more talent doing civilian crypto than classified crypto. The only way to think otherwise is to believe that tons of uber-geniuses are somehow found and snatched up before they ever publish any good work. Pretty unlikely, if you ask me.
(By "academic" positions, I mean those at universities or industrial research labs -- places that publish openly.)
Actually, in Z_p^* it's not even known how to efficiently map an element (y,g) to (z,h) where g and h are random generators and z and y have the same discrete logs with respect to their bases. So I strongly doubt that such a thing could be done between Z_p^* and elliptic curve groups, which have radically different structure.
I would doubt it. If your willing to spend a few million yearly keeping top tier researchers on your payroll, it's not all that hard.
Yes, it is. First, you radically underestimate the quality of researchers doing crypto in academia -- if you look at the people who were doing top-rate crypto work in graduate school and publishing at the best conferences, virtually all of them have taken academic positions after graduation. Second, a sizable fraction of those researchers are not US citizens. The rest of the world (Israel in particular) puts out some fantastically good cryptographers. None of them work for NSA.
To my knowledge, if (algorithm solving) DLP is "easy" over elliptic curves then the same algorithm can be (modified to be) used for integers "easily".
Maybe; I've never seen such a result, but this isn't my area of expertise. You'd need to be able to efficiently map a random generator and element of Z_p^* into a generator and element on the elliptic curve group, such that the elements have the same discrete log with respect to the bases. If there's a way, it's not at all straightforward.
You can show that this class of groups effectively maximises the difficulty of the Discrete Log Problem, and that's why the key sizes and computational efficiency is so much better.
Nobody has shown any such thing -- as far an anyone knows, DLP over elliptic curves is easy, but still hard over the integers mod primes.
However, the best *known* algorithms for solving DLP over elliptic curves are exponential-time (this may change, if more is learned about elliptic curves), while in the integers case they are subexponential-time. This makes a big difference in key lengths when you get down to implementations.
Everyone likes to bash how ugly Stata is on the outside. I like the way it looks, but I can see how others might not.
But, you really should walk through the "Student Street" area before making up your mind. It's pretty breathtaking: a big, open hallway with various corners of other buildings (made of brick, reflective aluminum, glass) sticking through the ceiling at odd angles. Walls painted with several strong, basic hues. Classrooms with cool polka-dotted echo-proof wood panels all over the walls (though these might give a headache after awhile). Lots of swooping stairwells that take you up to places where external walls from another building cut through the glass ceiling and continue all the way through the floor.
It's like a carnival funhouse. Soon to be inhabited by the carnies.
It still seems a bit too much of a coincidence, gnu going down exactly on the day on which (for example) debian.lcs.mit.edu was also unreachable, and the admins warned us about machines moving. Perhaps gnu.org was being hosted at LCS, but was transferred to a hosting company to coincide with the Stata move-in? Anybody know its hosting status from a few weeks ago?
gnu.org is down because the MIT CSAIL (Comp Sci and AI Lab), which houses servers such as gnu.org and debian mirrors, is moving into its new home in The Stata Center this week. Lots of important machines are being moved over from the old buildings to the new ones.
Remember, even though TLF has been proved, we still don't have the "simple proof" that Fermat himself discovered.
That's because he almost certainly didn't discover one.
Fermat was known for making some pretty bone-headed mistakes. Also, in his future writings he posed challenges to prove FLT for the case of n=3 or n=4, but never for general n>2. If he had found a truly elegant proof of the general case, and believed it was true, why not pose the general challenge?
Or, as I suspect, he's an invention of some half-inventive writer who's looking to run a nice Internet troll, maybe get a little play in the major news media.
I know the author of the story personally (we TA a class at MIT together). A month ago, he came in and said, "I just interviewed a guy who does tech support for the mafia! It was really fun."
So, I don't think this is a troll. It may be exaggerated (by the interviewee, or the author, or Wired), but it's not a total hoax either.
Re:Simson Garfinkel is a real author
on
Mafia Tech Support
·
· Score: 3, Informative
The facts may be changed to protect the "innocent," but the story is not a fabrication. I TA Ron Rivest's class with Simson, and he came in about a month ago and said, "So, I just interviewed a guy who works for the mob; it was really interesting..."
Of course, the interviewee could be a fraud. I don't know what he may have done to prove his story.
"D'ja ever notice how disk manufacturers are using 10^9 as 'giga' instead of 2^30? I remember back when we useta get a true 1024 multiplier for every step up the metric prefix ladder. 'Course, then every megabyte would set you back $20, but it was a full 1048576 bytes you were getting, and that was something you could count on. Nowadays it seems as if every swindler out there is trying to lowball his numbers, just to save a little magnetic coating. And don'tcha hate it how you have to get up seven times every night to go to the bathroom, and your joints ache from leaning down to pick up the toilet seat? And how nobody likes to listen to an old codger whine about insignificant crap like how big a megabyte really is? I'm a sad, lonely old man."
and they implemented an algorithm from Handbook of Applied Crypto for purposes of encryption with a value listed in the book that says explicitly "Do not use this for cryptographic purposes"
It was actually worse than this -- they used a Linear Congruential Generator, which is a very cheap method of generating "random" numbers. Those numbers might work well for simulations, but for cryptography they're totally predictable once you've seen just a couple of output values. Cryptography relies upon the unpredictability of random numbers for security, so LCGs should never be used for that purpose.
Under old-school English law, one peppercorn was the smallest amount of legal tender could legally make a contract over (sort of like $1 today).
PepperCorn was the original name of the business, but it ran into trademark problems, so they went with PepperCoin (which is more clever, anyway) instead.
Slashdot Mirror
Wow, he knocked a 'white doohickey' off his motherboard, walked around with it while his arm hairs were standing up straight from static electricity, and still expected the thing to work? What a chump. But not nearly as chumpy as someone who would do these things (i.e., me, with my first DIY system):
1. jammed a DIMM in backwards (this is hard -- the slot is asymmetric to avoid this very thing), turned the machine on, and quickly smelled the sweet smell of burning plastic as the DIMM holder melted, then tried to turn the machine off but forgot that you have to hold the power button down for several seconds, and stinking up the entire house before just pulling the damn plug...
2. vacuumed the dust out of the inside of the case while the machine was running, accidentally tapping the spinning CPU fan with the tip of the vacuum attachment, and snapping one of the fan blades off, making it spin out of control like a unbalanced centrifuge and making a horrible loud noise...
3. speculated that random machine crashes were being caused by a poorly-mounted heat sink, so removed the sink and turned the machine on, heard a loud "BEEEEEEP" and no start-up, then put his finger on the exposed die of the CPU to feel what was going on--OHDAMNIT'SHOTHOTHOTHOT, and enjoying the sweet smell of burning fingertip flesh...
What I was actually thinking of was never getting cryptanalysts get their hands on both the plaintext and ciphertext.
No, this is not a rule either. Any decent cryptosystem should assume that (and remain secure if) the analyst has access to as many plain/ciphertext pairs as he can handle.
Isn't the lowest possible hand a 7-high? You can't pick five cards lower than a seven without getting a pair or a straight.
In any respectable game of poker where low hand wins (Omaha Hi-Lo, or Lowball), a five-high (i.e., A2345) is the lowest possible hand. It doesn't count as a straight, and the ace counts as a low card.
I've done some old puzzles edited by Will Weng, and I've done a *ton* edited by Will Shortz. To me, there is no comparison. Will has really encouraged his constructors to have fun with their themes, and to creatively "break the rules" when it serves to create an "aha!" or "that's neat!" moment.
where there is a trick to how certain boxes can be filled in, for example a single box meaning 'ying' horizontally and 'yang' vertically
Awesome that you should mention this. The two authors of this puzzle are/were students in my research group (one graduated), and I think this is one of the best themes I've ever seen. I will mention your post to them!
Who is "we"? Civilian cryptographers?
If so, I repeat: if you look at the people who were doing top-rate crypto work in graduate school and publishing at the best conferences, virtually all of them have taken academic positions after graduation. When you take a reasonable look at these things, the far-and-away most likely conclusion is that there is vastly more talent doing civilian crypto than classified crypto. The only way to think otherwise is to believe that tons of uber-geniuses are somehow found and snatched up before they ever publish any good work. Pretty unlikely, if you ask me.
(By "academic" positions, I mean those at universities or industrial research labs -- places that publish openly.)
(Sorry to reply to my own post...)
Actually, in Z_p^* it's not even known how to efficiently map an element (y,g) to (z,h) where g and h are random generators and z and y have the same discrete logs with respect to their bases. So I strongly doubt that such a thing could be done between Z_p^* and elliptic curve groups, which have radically different structure.
I would doubt it. If your willing to spend a few million yearly keeping top tier researchers on your payroll, it's not all that hard.
Yes, it is. First, you radically underestimate the quality of researchers doing crypto in academia -- if you look at the people who were doing top-rate crypto work in graduate school and publishing at the best conferences, virtually all of them have taken academic positions after graduation. Second, a sizable fraction of those researchers are not US citizens. The rest of the world (Israel in particular) puts out some fantastically good cryptographers. None of them work for NSA.
To my knowledge, if (algorithm solving) DLP is "easy" over elliptic curves then the same algorithm can be (modified to be) used for integers "easily".
Maybe; I've never seen such a result, but this isn't my area of expertise. You'd need to be able to efficiently map a random generator and element of Z_p^* into a generator and element on the elliptic curve group, such that the elements have the same discrete log with respect to the bases. If there's a way, it's not at all straightforward.
You can show that this class of groups effectively maximises the difficulty of the Discrete Log Problem, and that's why the key sizes and computational efficiency is so much better.
Nobody has shown any such thing -- as far an anyone knows, DLP over elliptic curves is easy, but still hard over the integers mod primes.
However, the best *known* algorithms for solving DLP over elliptic curves are exponential-time (this may change, if more is learned about elliptic curves), while in the integers case they are subexponential-time. This makes a big difference in key lengths when you get down to implementations.
The NSA is the largest hirer of math majors.
Even if true, they still hire only a small, small fraction of all math and computer science PhDs.
Worldwide non-NSA talent vastly outweighs NSA talent, I would wager dollars to donuts.
gnu.org hasn't been hosted at MIT in years. Seriously.
Neat, I believe it. Too bad it doesn't explain gnu's downtime.
Everyone likes to bash how ugly Stata is on the outside. I like the way it looks, but I can see how others might not.
But, you really should walk through the "Student Street" area before making up your mind. It's pretty breathtaking: a big, open hallway with various corners of other buildings (made of brick, reflective aluminum, glass) sticking through the ceiling at odd angles. Walls painted with several strong, basic hues. Classrooms with cool polka-dotted echo-proof wood panels all over the walls (though these might give a headache after awhile). Lots of swooping stairwells that take you up to places where external walls from another building cut through the glass ceiling and continue all the way through the floor.
It's like a carnival funhouse. Soon to be inhabited by the carnies.
Nope. gnu.org is not hosted at MIT at all:
$ host gnu.org
gnu.org has address 199.232.76.164
Dunno why gnu.org was down
It still seems a bit too much of a coincidence, gnu going down exactly on the day on which (for example) debian.lcs.mit.edu was also unreachable, and the admins warned us about machines moving. Perhaps gnu.org was being hosted at LCS, but was transferred to a hosting company to coincide with the Stata move-in? Anybody know its hosting status from a few weeks ago?
No they don't, and no it isn't.
They go over the north pole or Canada, which *is* in the way.
gnu.org is down because the MIT CSAIL (Comp Sci and AI Lab), which houses servers such as gnu.org and debian mirrors, is moving into its new home in The Stata Center this week. Lots of important machines are being moved over from the old buildings to the new ones.
Remember, even though TLF has been proved, we still don't have the "simple proof" that Fermat himself discovered.
That's because he almost certainly didn't discover one.
Fermat was known for making some pretty bone-headed mistakes. Also, in his future writings he posed challenges to prove FLT for the case of n=3 or n=4, but never for general n>2. If he had found a truly elegant proof of the general case, and believed it was true, why not pose the general challenge?
Or, as I suspect, he's an invention of some half-inventive writer who's looking to run a nice Internet troll, maybe get a little play in the major news media.
I know the author of the story personally (we TA a class at MIT together). A month ago, he came in and said, "I just interviewed a guy who does tech support for the mafia! It was really fun."
So, I don't think this is a troll. It may be exaggerated (by the interviewee, or the author, or Wired), but it's not a total hoax either.
The facts may be changed to protect the "innocent," but the story is not a fabrication. I TA Ron Rivest's class with Simson, and he came in about a month ago and said, "So, I just interviewed a guy who works for the mob; it was really interesting..."
Of course, the interviewee could be a fraud. I don't know what he may have done to prove his story.
Or, to avoid these several steps, simply press and hold the shift key for a few seconds while inserting the CD into the drive.
This prevents the SbcpHid driver from being installed in the first place.
All true, except MIT is not Ivy League. It's in a league of its own. :)
In a nasally, whiny voice:
"D'ja ever notice how disk manufacturers are using 10^9 as 'giga' instead of 2^30? I remember back when we useta get a true 1024 multiplier for every step up the metric prefix ladder. 'Course, then every megabyte would set you back $20, but it was a full 1048576 bytes you were getting, and that was something you could count on. Nowadays it seems as if every swindler out there is trying to lowball his numbers, just to save a little magnetic coating. And don'tcha hate it how you have to get up seven times every night to go to the bathroom, and your joints ache from leaning down to pick up the toilet seat? And how nobody likes to listen to an old codger whine about insignificant crap like how big a megabyte really is? I'm a sad, lonely old man."
That's a pretty far-out story. I mean, really -- The Third Amendment? Nobody sues over that thing.
and they implemented an algorithm from Handbook of Applied Crypto for purposes of encryption with a value listed in the book that says explicitly "Do not use this for cryptographic purposes"
It was actually worse than this -- they used a Linear Congruential Generator, which is a very cheap method of generating "random" numbers. Those numbers might work well for simulations, but for cryptography they're totally predictable once you've seen just a couple of output values. Cryptography relies upon the unpredictability of random numbers for security, so LCGs should never be used for that purpose.
Under old-school English law, one peppercorn was the smallest amount of legal tender could legally make a contract over (sort of like $1 today).
PepperCorn was the original name of the business, but it ran into trademark problems, so they went with PepperCoin (which is more clever, anyway) instead.
(Sorry to reply to my own post, but this info is outdated.)
Now the pulp is sold to tissue makers. They bleach it white, make TP, and you wipe your arse and blow your nose with it.
Is this any way to treat our most valuable national secrets???