For many of us, The Terminator series introduced us to the potential dangers of artificial intelligence (AI). As Skynet's advanced AI became self-aware, it concluded that humanity was a threat to its existence and sprang into self-preservation mode, ultimately triggering a nuclear holocaust and deploying an army of Terminators to battle the resistance.
While this was purely fictional back in 1984, 35 years later, AI-powered threats are the new reality and raises the question: Are we headed for a Skynet-like future in which AI takes over the world? Perhaps we're not quite there yet, but the ingredients are all there and it could be a potential recipe for disaster.
As our understanding of AI progresses and evolves, AI attacks will get more sophisticated and continue to improve. Maturing open source machine learning tools like TensorFlow from Google and others will be used in malcode, distributing even more damaging botnets, viruses, worms, trojans, targeted phishing expeditions, and so on. Of particular concern is the combination of machine learning, automated facial recognition and huge amounts of data in recent dumps. This puts billions of people at risk of being compromised more than ever before.
One recent data dump is now raising alarm flags because it has the potential to affect millions of people. Known as Collections #1–5, well over 2 billion usernames and passwords were dumped onto the Dark Web. With data the foundation of AI, hackers can now carry out machine learning-based operations that leverage automated facial recognition and the information in Collections #1–5 to traverse social media networks and other sites to carry out automated spearphishing campaigns and a variety of other villainous exploits.
An AI populated with billions of email password pairs has a huge head start on leveraging evasive and powerful attack tools such as DeepLocker and Social Mapper. Consider the kill chain of shared credentials between corporate and personal emails. That's a very soft target for the Terminator of malware. Even if only 1% of the passwords in the "Collections" are still accurate and shared across accounts, that is well over 20 million vulnerable victims. From statistical analysis, we know the rate is far higher than that.
So, how bad could it get? Realistically, a mass collective hive of botnets with knowledge of credentials, email, facial recognition, and social networks could make AI phishing lures that will be make email unusable. Theoretically, with Collections #1–5 at its disposal, Skynet could now take over the world.
Which leads us to the need for a Resistance. Fortunately, Skynet does not exist… at least, not that we know of. But it will take a lot more than John Connor to win the AI war with cybercriminals. It will take a global coalition of brilliant minds and organizations from the private and public sectors fighting fire with fire, deploying AI-based security solutions that can keep pace, outmaneuver, and outthink these AI-powered attacks. The US Department of Defense echoed this sentiment in a recently unveiled summary of its official artificial intelligence strategy:
We cannot succeed alone; this undertaking requires the skill and commitment of those in government, close collaboration with academia and non-traditional centers of innovation in the commercial sector, and strong cohesion among international allies and partners. We must learn from others to help us achieve the fullest understanding of the potential of AI, and we must lead in responsibly developing and using these powerful technologies, in accordance with the law and our values.
Perhaps the late Stephen Hawking said it best: "Unless we learn how to prepare for, and avoid, the potential risks, AI could be the worst event in the history of our civilization."
Or as the Terminator might say: "Hasta la vista, baby."
Join Dark Reading LIVE for two cybersecurity summits at Interop 2019. Learn from the industry's most knowledgeable IT security experts. Check out the Interop agenda here.