LAS VEGAS, NEVADA — Artificial Intelligence (AI) technology has gradually infused itself into our lives, making its potential threats something that we cannot ignore.
As a global platform for exploring AI and information security, GeekPwn focused this year on AI security at DEFCON, the world’s largest yearly hacking convention, which just ended in Las Vegas.
“GeekPwn hopes to establish a bridge between white hat hackers and other technical fields, and let white hats explore wider horizon fields by adopting hacker-like thinking,” Wang Qi, founder and CEO of The KEEN Team, a white-hat hacking collective in China that organizes annual GeekPwn competitions, told Xinhua.
During DEFCON, widely regarded as the birthplace of the competition of cyber-attack and defense technology, GeekPwn debuted the first global CAAD CTF (Competition of Adversarial Attacks and Defense, Capture the Flag) event.
“It is the first CTF in machine learning field, surpassing the limitations of A&D of traditional cybersecurity competitions,” Wang told Xinhua. “For the first time, participants can view the complexities of conducting attack and defense in the AI world.”
The contest utilized a randomly on-site matching mechanism, which placed six top AI security research teams to create adversarial samples to defraud and attack other teams’ defense algorithms.
The event was joined by experts from institutes such as Tsinghua University, UC Berkeley, Cornell University, Yale University, JD.com and others.
After a close contest, the TSAIL team from Tsinghua University emerged champion of the CAAD CTF contest.
In addition to the AI attack and defense battles, GeekPwn2018 in Las Vegas also brought 10 latest research results on AI security and adversarial samples.
The topics covered included autonomous vehicles, speech recognition systems, anti-virus software and neural networks, which are highly watched by AI entrepreneurs, practitioners and scholars.
Jia Yunhan, a senior researcher from Baidu Security Lab, brought a demo of “Blind Eye Magic” from the automatic driving industry, revealing how the use of adversarial machine learning technology can possibly deceive automotive radar systems.
Rohan Taori and Amog Kamsetty, graduate students from the University of California at Berkeley, shared the topic of directional “black box” attack analysis for audio systems.
Bo Shi, director at the Tencent Security Yunding Lab, shared contents of dogfight weapon GAN, explaining how malware and viruses fight against anti-virus (AV) software.
“The industry is benefiting from the growing need of cyber security solutions and services in the defense,” Wang told Xinhua. “Security risks are always making the world realize the importance of prevention and value of white hat hackers.”