Hackers Target AI To Increase Energy Consumption And Slow Systems

BACKGROUND:

A new type of attack could increase the energy consumption of AI systems, according to research undertaken at Cornell University. Similarly to DDoS (distributed denial-of-service) attacks on the internet seeking to clog up a network and make it unusable, the new attack forces a deep neural network to tie up more computational resources than necessary and slow down its “thinking” process. The slowdown attack targets a type of AI called an input-adaptive multi-exit neural network, which can be deployed on small devices like smartphones and smart speakers. Theoretically, if an attacker had full information about the neural network, they would be able to max out its energy draw.

Experts Comments

May 10, 2021
Jake Moore
Cybersecurity Specialist
ESET

With advanced computing comes advanced attacking. Such threats are often overshadowed by the emerging technology itself, but they could potentially take out networks with attacks that haven’t even been discovered by the developers.

 

True testing of newer technologies requires ethical hackers to think like criminal hackers. Without such measures, we stand the chance of seeing adversaries continue to take advantage and outstrip legacy security.

What do you think of the topic? Do you agree with expert(s) or share your expert opinion below.
Be part of our growing Information Security Expert Community (1000+), please register here.