Follow Slashdot stories on Twitter


Forgot your password?
Check out the new SourceForge HTML5 internet speed test! No Flash necessary and runs on all devices. ×

Submission + - Microsoft Sees Over 10 Million Cyberattacks per Day on Its Online Infrastructure

An anonymous reader writes: Microsoft's user identity management systems, made up by Microsoft Account (formerly Live ID, for home users) and Azure Active Directory (for its cloud/corporate services), see over 13 billion user logins per day, with 1.3 billion for AAD. The company says that over 10 million (per day) of these login attempts are cyber-attacks, which the company is able to detect.

This information comes via Microsoft's most recent Security Intelligence Report, which also reveals details about a new cyber-espionage group named Platinum and that hackers are still using the same vulnerability (CVE-2010-2568) even today, which was used in the Stuxnet attacks.

Submission + - A $6 Bot That Sinkholes Telemarketers and Robocalls 1

Trailrunner7 writes: A bot that started as a way for one man to annoy and frustrate telemarketers and robocallers has now developed into a subscription service for consumers and businesses who have had enough of the unsolicited and sometimes fraudulent calls.

The Jolly Roger Telephone Co. is the creation of Roger Anderson, a phone industry veteran who built it for his own use originally. The concept is simple and ruthlessly effective for preventing robocalls from getting through. The system sits in front of a landline and when a robocall is detected, it will respond with a phrase to get a human on the line or make the robocalled think it’s talking to a real person. Anderson had the system on his home lines and it worked well enough that he eventually posted instructions for others to send telemarketers to his bot.

He then built more bots with different abilities and says that the bots have answered nearly 200,000 calls so far. His Jolly Roger bot got quite a lot of media attention a couple months ago and so Anderson decided to put together a subscription service to allow consumers and businesses both to send unwanted calls through the system. The subscription price is $6 per year for consumers.

Submission + - What's the smallest biggest number you can think of?

serviscope_minor writes: If you think exponentials, factorials or even Ackermann's function grow fast, then you're thinking too small. For truly huge, but well defined, numbers, you need to enter the realm of non computability.

The Busy Beaver function BB(n) is the largest number of steps that an n state Turing machine will run for when fed with a blank tape excluding non halting programs. It grows faster than any computable series but starts off as the rather pedestrian 1, 6, 21, 107. By BB(7) it reaches at least 10^10^10^10^10^7 and at some point becomes non computable. It must be non computable because if it wasn't, you could run a program for BB(N+extra states needed to encode the initial tape state)+1 steps, and if it gets that far then you know it never halts, so you've solved the Halting Problem. So, at some point it must transition from numbers that can be computed to ones that can't be.

And now there's some new and rather interesting insight into that which essentially reduces the problem to code golf or the International Obfuscated Turing Code Contest (as if there is any other sort). Imagine you have an axiomatic system, say ZFC (which underlies almost all of modern maths), and you know you can't prove it's consistent (you can't). If you write a program that systematically evaluates and tests hypothesis based on the axioms, you can't prove it will halt or not since that's equivalent to proving consistency.

This insight and first upper bound is the program proving that BB(7918) is noncomputable comes from this new paper. It turns out that writing a ZFC axiom evaluator directly in a Turing machine is rather tricky and long winded, so the authors wrote a small interpreter for a higher level language then wrote the axiom evaluator in that. Now finding a smaller uncomputably larger number is a question of writing even smaller programs which attempt to compute undecidable things. Think you can do better? A good starting point would probably be the existing code on github.

(I hope I've got the explanation at least half way right!)

Slashdot Top Deals

One good suit is worth a thousand resumes.