Designed to help protect artificial intelligence (AI) systems IBM has released a security library into the open-source community. The toolbox, released to the open-source community, aims to become a repository and source of information on threats to our current and future AI solutions.
To confuse AI, undetectable alterations in content including images, video, and audio recordings can be crafted. IBM believes, that for this, one does not need a deep knowledge of the AI or DNN to exploit weaknesses in AI systems.
These changes can be small in size but cast huge security concerns. They impact the performance of AI systems like prompting them to make a choice which we would deem malicious.
Aimed to combat so-called “Adversarial AI”, the toolbox records threat data as well as assist developers in creating, bench-marking, and deploying practical defence systems for real-world AI. IBM shared that this research looks at the best ways to defend the AI systems before the bad guys attack.
By introducing the toolkit to the open-source community, others may also become inspired enough to create solutions before Adversarial AI becomes a true threat. The toolbox also includes a library, interfaces, and metrics which will help developers begin to create cyber security solutions for this emerging field.
“Considering, tools didn’t provide the defences needed to protect AI systems, this is the first and only AI library that contains attacks, defences, and benchmarks to implement improved security,” the company says. Read more…