Cybertek Defense Attacks, Threats, and Vulnerabilities Adversarial AI & Supply Chain Attacks

Adversarial AI & Supply Chain Attacks

0 Comments 5:57 pm

Part 2 – 8: Terms and Definitions for Adversarial AI & Supply Chain Attacks

TermDefinition
Adversarial AITechnique that attempts to fool AI programs with deceptive data
Supply Chain AttacksDelivering viruses or other malicious software via a supplier or vendor

AI (Artificial Intelligence) has come along way over the past decade. We use AI with a lot of our technology to help improve services and functionality with how users interact with devices and software. AI is meant as a learning process for computers to understand how humans use technology and what they prefer, while a heavy focus on categorizing what the end user wants to see. An example of how we use AI today could be advertisements. If you use Google and make some searches on clothing you want to buy, AI that is integrated into the search engine you use will promote clothing based on what you are currently looking at. The last thing you want to see in an Ad banner is something about how to fix cars when the search is for clothing. The meaning of adversarial AI comes when the AI is meant to be something for good, but the end user manages to manipulate the AI process. An example of this could be a captcha code for a website. When the captcha asks for what images look like a “car”, you could select everything else but a car. By doing this, you are letting the database know the car was not present and you selected a stop sign instead. While this process would take a long time to manipulate the program, it is possible to change the AI to learn the wrong things.

Supply Chain Attacks are similar but a different malware attack. Supply chain attacks target the vendors instead of the end users. If the attacker can infect the vendor for a software their clients purchase, every download that occurs will have the malware installed with the software. An example of this could be a camera device that is installed in every phone could have malware implanted on the chip to allow attackers to easily access the camera to turn it on/off at free will. The vendor may not know of the malware during the supply chain process when the cameras are installed, and would require software updates to fix the issue post release.