Q&A The Truth About Machine Learning In Cybersecurity: Defense

Discussion in 'General Security Discussions' started by Umbra, Dec 22, 2017.


Do you agree?

  1. Yes

    6 vote(s)
  2. No

    2 vote(s)
  1. Umbra

    Umbra From Emsisoft

    May 16, 2011
    Community manager
    Vietnam & France
    Windows 10
    Read more here


    How Cybercriminals can use Machine Learning (2nd part of the series)

    Read more here
    BryanB, Tsiehshi, Andy Ful and 10 others like this.
  2. Opcode

    Opcode Level 18
    Content Creator

    Aug 17, 2017
    Windows 10
    Machine Learning is good but it is flawed just like Ai/even checksum hash detection. I mean look at finger-print passwords, it is flawed with theft of someones prints extracted from a coffee table or glass/bowl. There is no silver bullet, never has been and never will be. We are the best line of defence, but we are neither a silver bullet as we aren't perfect. Humans have never been perfect and never will. And anything we create will follow this. Meaning any Ai systems we develop, they won't be perfect... Any Ai systems made by an Ai system won't be perfect, because the Ai system which made the AI systems Ai system (etc etc) will have originated from a human, which isn't perfect

    Only way a silver bullet could ever theoretically exist is technology from aliens, if they exist. But still I doubt aliens are perfect either.

    No matter what you got, someone out there has something better. Whatever you know, someone out there knows twice as much. And it goes on forever.
    Tsiehshi, BryanB, Andy Ful and 12 others like this.
  3. Prorootect

    Prorootect Level 46

    Nov 5, 2011
    0wN3D by my cat!
    Bookmarked: Peerlyst community managers picks for the top security posts, discussions and insights | Peerlyst: Peerlyst

    Thank you Umbra!
    BryanB, Umbra, Opcode and 2 others like this.
  4. Lockdown

    Lockdown From AppGuard

    Oct 24, 2016
    AppGuard LLC Virginia, U.S.
    IT security is pretty much a reactionary endeavor.

    There are those who are "sensitive" to any mention of the weaknesses of Ai\Machine Learning. I guess they interpret it as some kind of product bashing.
    Tsiehshi, Opcode and Vasudev like this.
  5. jerzy601

    jerzy601 Level 13

    Jun 20, 2011
    Windows 8.1
    I completely agree with this statement, nothing more to add.
    this is true.
    BryanB, Opcode and Behold Eck like this.
  6. Andy Ful

    Andy Ful Level 22

    Dec 23, 2014
    Windows 10
    Any good thing may be used for a malicious purpose, and vice versa. You can use viruses to heal people and exploit love to hurt somebody. AI is simply another feature that people can/will use both for the good and the bad.:coffee:
    BryanB and Opcode like this.
  7. Kubla

    Kubla Level 2

    Jan 22, 2017
    United States
    It will be a clash anti-malware AI vs AI malware we just have to hope the anti side can stay one step ahead of the AI malware.

    The biggest problem I see is that those developing AI malware will be government backed (taxpayer financed) tools to be used for spying and just like other spy tools these AI malware's will no doubt be proliferated and used for even more nefarious endeavors. In a nutshell the worlds tax paying citizenry are paying to develop AI to spy on them that will be used to exploit them, rob them or worse and then forcing them to pay for anti-malware security to protect them against what they reluctantly paid for in the first place.

    Pitting current anti-malware companies against multi-billion dollar backed AI malware tools is going to be asking a lot.
    BryanB and Andy Ful like this.
  8. Slyguy

    Slyguy Level 22

    Jan 27, 2017
    Fortinet Engineer
    Other OS
    #8 Slyguy, Dec 24, 2017
    Last edited: Dec 24, 2017
    I have a theory on that.. Aliens would likely be so advanced that security, locks and passwords would probably not even be a concept they understood or could be made to understand. Probably thousands of years in their past, a sort of ancient way of doing things.. I'm guessing they'd probably have ventured into areas of telepathy and other things (machine actualized or otherwise) that would likely render such things nonsense in a very modern existence.

    Barring that.. Machine learning is probably going to end up as just part of other layers. Fortinet uses it in their pre-scanning of incoming files for the FortiSandbox. That's about it in their offerings and it is done to 'speed' quantification of incoming data at a rate so fast most people can't discern it is even scanning it. I don't believe it's a panacea like some people (and companies) claim, but just another tool.
    Andy Ful, BryanB and Opcode like this.