Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!


Forgot your password?
AI Japan Security

Japan Leads Push For AI-Based Anti-Cyberattack Solutions (nikkei.com) 34

An anonymous reader writes: Japanese firms NTT Communications and SoftBank are working to develop new artificial intelligence (AI) platforms, offering cyber-attack protection services to their customers. Up until recently, AI-based security systems were only used for certain scenarios, in online fraud detection for example. The new offerings will be the first commercially-available platforms of their type for use in a wide range of applications.
This discussion has been archived. No new comments can be posted.

Japan Leads Push For AI-Based Anti-Cyberattack Solutions

Comments Filter:
  • No dinochrome brigade...for now.

  • Clarify... (Score:5, Insightful)

    by fuzzyfuzzyfungus ( 1223518 ) on Tuesday October 13, 2015 @05:55AM (#50716111) Journal
    This seems like it could well be a viable thing; but 'AI-based' is serious weasel-word territory: is a Baysian spam filter an "AI-based anti-spam solution"? It's hard to argue with the notion that identifying anomalous activity in large volumes of traffic is a problem that might be amenable to statistical methods and assorted heuristics; but what exactly qualifies or disqualifies something for 'AI-based', 'deep learning', and similar buzzwords?
    • by Rinikusu ( 28164 ) on Tuesday October 13, 2015 @06:13AM (#50716165)

      In their case, when it detects a cyber attack, it signals the giant mecha robot station in LEO to launch powered cybernetic suits that enter the earths atmosphere and land on top of the attack command and control. These suits then rip the roofs off the houses/offices of the attackers and pulls their pasty, fat asses out of their chairs and slice them to pieces using energy blade weapons. Unless the attackers are female, at which point they are returned to the giant mecha robot station to be brainwashed and trained as cybernetic suit pilots with psychic powers and fucked up emotional issues of identity, purpose, and a weird love 2+n(angle).

    • One of the aspects of AI/machine-learning is deep down it's only one algorithm. Just like a human mind/brain, it learns any skill-- the underlying brain structure/neurons is the same. It can do things like object recognition in images(already available in google my-photos), facial recognition, voice recognition, language translation etc.

      So to call something as AI it must just take in huge amounts of data and figure out a structure/classification in the data all by itself. It's like you are given 100 pic

    • is a Baysian spam filter an "AI-based anti-spam solution"?

      If the probabilities are hard coded, then no. If it learns and updates itself, then yes, that is AI. AI is determined by behavior, not mechanism.

      what exactly qualifies or disqualifies something for 'AI-based'

      Programs that learn and adapt are AI.

      'deep learning', and similar buzzwords?

      "Deep learning" refers to neural nets that have more than one hidden layer and are thus "deep". They are usually based on RBMs [wikipedia.org]. "Deep learning" has a very specific meaning, and is not a "buzzword".

  • This is great. I sadly missed out (was too young) for the whole Y2K money fest, but I saw the new BMWs and expensive vacations that many of my older computing peers undertook as a result of the exorbitant amounts of money they charged to review old source code. I also knew enough then to realise that about 95% of what 'experts' were saying on the TV was nothing more than fear mongering.

    Anyway with this cyber-warfare stuff kicking in there should be no shortage of cushy contract work for decades to come. The

    • The Beltway Bandits are several years ahead of you. They all glommed onto "cybersecurity" as the next big cash cow following the fading of "counterterrorism." 2009-2011 was a really good time to own a small contracting firm that specialized in cybersecurity, because Northrop, Lockheed, General Dynamics, etc, all went on a buyout spree trying to get both contracts and expertise. I only wish I'd had the foresight/connections at the time..
  • Skynet ? (Score:5, Informative)

    by jcdr ( 178250 ) on Tuesday October 13, 2015 @06:16AM (#50716175)

    At least a step in that direction.

  • by nickweller ( 4108905 ) on Tuesday October 13, 2015 @06:28AM (#50716209)
    Until these innovators can design a computer that can tell the difference between data and code and don't execute code downloaded over the Internet, then such solutions are just so much snake oil.
  • Instead of assuming that the programs you run can be trusted, flip the assumption, and a lot of "cyber security | cyber war" crap goes away. This can be fixed, folks.

  • And no one notices? Relying on a machine to do all the work sounds a bit like leaving a dog to guard against other dogs. It might work or they might end up sniffing each others balls.

  • This will NOT end well.....

    • Don't worry - they're planning to partner with the Japanese robotics company, Cyberdyne ( https://en.wikipedia.org/wiki/... [wikipedia.org]. ) to make a powerful AI system that will protect their customers on the internet at first, and possibly even later in the physical world.

      It's all perfectly safe....
  • So maybe we can hack the program to attack, I don't know, Pearl Harbor? Better be some good code running on their own hardware by people who eat nigiri for breakfast. When it works it would be awesome, but when that update goes wrong it could be hell.

Been Transferred Lately?