The GeekPwn 2017 Carnival Offers Prize of 5 Million RMB with new AI Security Session

SHANGHAI, June 20, 2017 /PRNewswire/ -- Global security experts are invited to participate the GeekPwn 2017 Carnival in Silicon Valley (US) and Shanghai (China) starting from October 24. GeekPwn, the first worldwide security geek contest for smart life, offers 5 million RMB as prize pool for successful "PWNing" with talent, break-limitation minds and codes.

GeekPwn, a stage for white-hat hackers "PWNing " smart devices, finding and fixing system vulnerabilities and protecting smart life, set a special AI (artificial intelligence) Security session for the first time in 2017. It includes AI PWN and PWN AI submissions, which finally improve people's smart life in the AI age.

AI Security Becomes a Necessity

After noticing AI security issues, GeekPwn encourages contestants to test and improve AI's mechanism on security and analysis, which finally ensures AI's development in a healthy and sustainable way.

It brings lots of uncertainty and concerns of human being's future if AI is maliciously controlled and mislead. For example, some products were mislead and cheated recently on picture recognition and classification, which was used by Internet crimes to spread violent, porn and terrorism pictures online.

GeekPwn has started research on "antagonistic machining learning practice" and recruited and touched top-level AI security talents since the end of 2015, when AI-related risks were exposed initially. During the GeekPwn 2016 Silicon Valley session, OpenAI scientist Ian Goodfellow, the inventor of GAN, shared the latest achievement - the "antagonistic imaging" can cheat AI easily on picture recognition in the real world. The fact of that AI can be cheated triggered GeekPwn's attention, because AI's error and imperfection become loopholes used by hackers.

PWN AI and AI PWN

The special AI Security session includes two types of submissions as PWN AI and AI PWN.

PWN AI exploits vulnerabilities in the AI systems to make them stop working or make wrong decisions, covering all public AI services, products, libraries and frameworks. The target areas include computer vision, voice recognition, natural language processing, autonomous driving, malware detection and etc. For example, hackers with PWN AI are able to use any faces to unlock a phone with facial recognition and exploit vulnerabilities in autonomous driving system and make the system unable to detect some specific obstacles.

AI PWN takes AI, such as algorisms in computer vision, voice recognition, natural language processing, autonomous driving, etc., as primary or assistant tool in hacking/PWNing targets. For example, AI PWN includes using AI for speech synthesis to simulate the target people's voice and pass target authentication system with high probability; using AI to determine the hand actions from video clips to identify the password input with high correct recognition rate and using AI to identify complex machine codes with high correct recognition rate.

"PWN Everything" to Protect the World

Besides the new AI Security Session, GeekPwn continues its DNA of "PWN Everything" as setting no limit on PWN targets. Smart devices, IoT (Internet of Things) products in public markets are all acceptable PWN targets. Contestants can PWN all pubic devices, through getting system control, accessing private data or breaking through original security mechanisms in reasonable attack conditions.

Unlock smart phones with noses, remote hijack any one of the world's communications and find the backbone of network equipment vulnerabilities that impact hundreds of millions of Internet users - those PWNing cases already existed in the previous GeekPwn events. GeekPwn offers white hat hackers and top security experts a unique stage to break the limitation of minds and show great power. Besides showing talent, white hat hackers in fact help several hundred firms to fix security loopholes and vulnerabilities, which finally brings safer smart life.

www.geekpwn.org

PWN AI

Targets

For all public AI Services, Products, Libraries, Frameworks, if you can exploit vulnerabilities to make the AI system or component stop working, or lead the AI system or component make wrong decisions, please register. The target areas include Computer Vision, Voice Recognition, Natural Language Processing, Autonomous Driving, Malware Detection, etc. The target AI frameworks include mainstream frameworks like TensorFlow, TorchNet, Caffe, etc.

Examples

Use any face to unlock a phone with facial recognition.

Exploit vulnerabilities in autonomous driving system, make the system unable to detect some specific obstacles.

Exploit vulnerabilities in AI framework, make deployed AI system stop working in some specific situations.

AI PWN

Scenario

Contestant takes AI (Various algorisms in Computer Vision, Voice Recognition, Natural Language Processing, Autonomous Driving, etc.) as primary or assistant method in hacking process to break the limit of target system. Therefore the original functions of target system stop working, or information leaked.

Examples

Using AI method for speech synthesis, simulate the target people's voice and pass target authentication system with high probability.

Using AI method to determine the hand actions from video clips to identify the password input with high correct recognition rate.

Using AI method to identify complex CAPTCHA with high correct recognition rate.

About GeekPwn 2017

GeekPwn is an international security community focusing in smart life. GeekPwn is held by Keen every May 12th and Oct 24th to provide chances for security geeks to show their talents.

From IoT, Smart devices to AI services, any successful compromises of security restrictions have chances to be accepted. New match types, geek shows and PWN of AI are available on GeekPwn 2017 Carnival held in Silicon Valley (US) and Shanghai (China).

About GeekPwn lab

GeekPwn lab is the security research team of Keen Cloud Tech. The team focuses on helping worldwide leading software manufacturers, which have adopted advanced security engineering methodologies, to discover and fix security vulnerabilities. In the past years, GeekPwn lab has discovered and reported hundreds of high-risk vulnerabilities to Microsoft, Apple, Google, etc. GeekPwn lab is currently focusing on the security research of cloud computing security and mobile security.

SOURCE KEEN