Stoecklin, Principal RSM & Manager, CCSI at IBM Research demonstrated to TechRepublic's Dan Patterson just how new artificial intelligence-powered facial recognition technology can trigger malware lurking within common applications. Stoecklin: What we show in this proof of concept is AI-powered malware through a distribution channel, which is using unsuspicious, innocent-looking application. We use for this purpose a videoconferencing application that we call Talk. The user is opening the application from his download and it is running. We have the sign-in screen. Now, the application can be used as if it was a normal application.

TechRepublic's Dan Patterson sat down with Jiyoung Jang, Research Scientist, CCSI Group at IBM Research, Marc Ph. The researchers have discovered invasive and targeted artificial intelligence-powered cyber-attacks triggered by geolocation and facial recognition. The following is an edited transcript of the conversation. Stoecklin, and Dhilung Kirat: IBM Research, and specifically our team, has a long tradition in analyzing technology shifts out there, and how they impact the security landscape out there. Then we understand how to counter these attacks, and how to give recommendations to organizations.

What if machine learning models, much like photographs, movies, music, and manuscripts, could be watermarked nearly imperceptibly to denote ownership, stop intellectual property thieves in their tracks, and prevent attackers from compromising their integrity? Thanks to IBM's new patent-pending process, they now can be. In a phone conversation with analysts this week Marc Stoecklin, IBM's manager of Cognitive Cybersecurity Intelligence, detailed the work of several IBM researchers who've been busy trying to find new ways to embed unique identifiers, or watermarks to you and I, into neural networks. Their concept was recently presented at the ACM Asia Conference on Computer and Communications Security (ASIACCS) 2018 in Korea, and might be deployed within IBM or make its way into a client-facing product in the very near future. "For the first time, we have a [robust] way to prove that someone has stolen an [AI] model," Stoecklin said.

LAS VEGAS--IBM will detail at Black Hat USA here on Aug. 8 a new class of attacks dubbed DeepLocker that uses artificial intelligence to bypass cyber-security protections. With DeepLocker, IBM researchers will demonstrate an evasive attack vector that has been developed as a proof of concept. According to IBM, DeepLocker can be used to keep ransomware or other malware hidden from traditional security tools. IBM's goal with the presentation is not to promote fear about AI, but rather to help organizations start to think about how attackers can use AI and how to minimize risks. "DeepLocker malware is fundamentally different from any other malware we are aware of. It uses AI to hide a malicious application in benign payloads," Marc Ph.

The creation and development of a robust neural network is a labor-intensive and time consuming endeavor. That's why a team of IBM researchers recently developed a way for AI developers to protect their intellectual property. Much like digital watermarking, it embeds information into a network than can then be triggered for identification purposes. If you've spent hundreds of hours developing and training AI models, and someone decides to exploit your hard work, IBM's new technique will allow you to prove that the models are yours. IBM's method involves embedding specific information within deep learning models and then detecting them by feeding the neural network an image that triggers an abnormal response.