Shulou(Shulou.com)06/01 Report--

The biggest shock to me this year is the negative problems caused by the AI robot. 4 billion harassing phone calls are broadcast, 5000 a day, 4 billion a year, more than 5000 devices are active per day, more than 30, 000 devices are in use, and more than 600 million users' privacy is compromised.

1. AI is not guilty, but he is guilty.

AI robot instead of manual, always polite and never angry and tireless voice, deserves to become the main force of harassing phone calls. The scheme of AI robots instead of manual harassment calls has formed a complete black industry chain of "AI calls + AI evades supervision + AI collects personal privacy". When people use the ability of AI to do security and improve work efficiency, AI has become a double-edged sword. Just like on RSAC this year, Google and Macfee both mentioned the negative problems caused by the application of AI technology.

In view of the security risks of AI itself, the previous article has combed the "RSAC2019 Security Technology Hot spots Serial (1) the dark side of machine learning".

Second, use the "shield of AI" against the "spear of AI"

The main purpose of this article is to share how to better use the power of AI to deal with the problems caused by the malicious exploitation of AI. For example, in the underground industry, anti-fraud, automatic analysis, automatic advertisement interception, reverse verification client, automatic monitoring of sensitive information, automatic shielding of bad Internet information and so on. In fact, we can see that the ability of AI to fight AI may be the ultimate solution of security evolution in the future.

Making use of the rest time, the author summarizes four kinds of application directions of using AI against AI at present:

The first category: AI anti-fraud platform:

1. Anti-fraud platform of financial institutions

2. Automatic analysis and evaluation platform of network security insurance.

The second category: AI inspection robot:

1. Automatic interception of online and mobile ecosystems, websites and other advertisements

2. Social media event monitoring and brand protection based on AI.

3. AI detects and fights all kinds of machine programs\ robots

The third category: AI content discovery and privacy protection

1. Content security of online browsing based on AI

2. Based on AI+ biometrics + personal privacy protection

Category IV: AI*** "Blitzkrieg" confrontation

1. AI promotes automation and shortens preparation time.

2. Use AI ability to automate orchestration and response to improve defense efficiency.

Third, the analysis of the AI direction of the start-up enterprise in Israel.

1. Unbotify,AI detection robot (established in 2015 and acquired by mobile monitoring and fraud prevention company Adjust at the end of January this year)

Official website: http://www.unbotify.com/

Core business direction: robot detection, which can distinguish between real people and robots by analyzing the human behavior patterns in the user flow of websites and mobile applications, so as to prevent the occurrence of digital fraud. Using machine learning and deep learning technology, Unbotify extracts hundreds of features from project data such as device direction (device orientation), touch event (touch events) and pressure sensitivity for detection and analysis, so as to identify cheating robots. This approach enables marketers to identify fraudulent robot activities without confusing legitimate human traffic, thereby ensuring interaction between application publishers and human audiences.

What acquirer Adjust is engaged in: mobile data monitoring and fraud prevention platforms that provide global mobile app marketers with high-quality analysis, monitoring and fraud prevention solutions to help them make marketing decisions more accurately and quickly.

2. NsKnox, AI Financial Anti-Fraud (established in early 2016 and completed a round of financing of US $15 million in February this year, led by Microsoft M12 Venture Fund and Viola Ventures)

Core business direction: financial anti-fraud platform. Protect corporate payments from multiple threats of fraud and data manipulation. NsKnox's solution leverages what the company calls Cooperative Network Security (CCS), which combines encryption with data from multiple organizations.

3. ChameleonX, AI online Anti-Robot Advertising (established in 2015, received US $1.9 million in seed wheel funding in 16 years, led by TrueVentures)

Official website: https://chameleonx.com/

Core business direction: protect online and mobile ecosystems from ad injection, robots and malicious advertising with machine learning capabilities.

4. Personal image privacy protection of DID,AI (established in 2017, US $4 million in seed round financing was completed in January 2018, led by Pitango Venture Capital)

Official website: https://www.deidentification.co/

Core business direction: companies with anti-image recognition technology to protect images from unauthorized automatic facial recognition systems, which contributes to the privacy of personal biological information. The deep learning solution incorporates a data protection technology called "de-identification", which can generate photos and videos that cannot be recognized by the algorithm, but remain similar to real faces to ensure that personal privacy and identity information are not maliciously read by facial recognition technology.

5. Cyabra-AI Internet Water Army Cleaner (established in 2017, the amount of seed round financing is unknown)

Core business direction: to reduce the adverse impact of online water army and false news on consumer brands and the government. Several of the founders of Cyabra have worked in the Israeli military and business intelligence companies. Cyabra is positioned to protect customers from the loss of an army of hooligans. Identify the possibility that each account is a false account through its basic information, photos, content and style of writing. In addition, because fake accounts often post the same / similar content across platforms, Cyabra obtains the real fake account database through various means and expands the fake account base based on the similarity of accounts on different social platforms. For suspected or confirmed false accounts, the company will provide customers with early prediction, early warning and next behavior prediction services. Cyabra has strong web crawler ability, user profile accumulation and corpus emotion analysis technology, and continues to climb the slope in natural language processing technology. Network public opinion belongs to the extremely difficult field of natural language processing applications, which is due to the relatively weak standardization of network language.

6. Shieldren, AI protect children's online health (established in 2017, the amount of seed round financing is unknown)

Core business direction: use machine learning and deep learning to monitor children's browsing on the Internet and protect children from browsing healthy websites.

7. Cyberwrite, online insurance automatic evaluation platform (established in 2016, the amount of seed round financing is unknown)

Official website: https://cyberwrite.com/

Core business direction: automatic evaluation of network insurance amount, provide automatic evaluation platform and services.

There are four main categories of network insurance coverage:

1) Technical errors and errors (Errorsand Omissions)

For example, the economic losses caused by the delay in product development and errors in product integration in IT enterprises

2) copyright, trademark and other intellectual property rights (MediaLiability)

For example, intellectual property disputes with other enterprises require legal fees, compensation, etc.

3) Network and information security

For example, the losses caused by trade secrets stored by enterprises or user data disclosure, data loss, virus transmission and extortion.

4) Private data

It is also the content of the data level, private data mainly refers to mistakes in the real environment, such as the loss of notebooks with sensitive data, employees will send wrong e-mails with customer information, and so on.

IV. Thinking and summary on the security direction of AI against AI

1. AI ability has become the core competence against machine learning / robot.

2. It has become a cognitive necessity for security practitioners to use AI against AI.

3. The automatic analysis and evaluation platform of network insurance is valued by security start-up enterprises.

4. AI is promising in the field of privacy protection of personal biometric information.

5. Using AI to ensure children's access to the Internet belongs to the innovative application field of AI.

6. AI is widely used in anti-fraud and advertisement interception in the field of financial and insurance Internet platform.

7. AI ability to achieve automatic early warning and response and policy scheduling, and then to combat AI automation * *

Fifth, the ultimate conjecture: in the future, the ultimate evolutionary ability of security is the confrontation between AI.

Although we are only in the stage of weak artificial intelligence, AI machine learning and automation have become weapons in the hands of hackers. Security confrontation becomes more and more complex because of the fusion of big data and artificial intelligence scenes. The ability of security experts should not be limited by massive security big data. Security experts released through AI tools can do more valuable work. Therefore, the author's immature guess: according to the "2008 Law", in the future, AI tools will spend 20% of the time to complete 80% of the work of security practitioners, while 80% of security experts will devote themselves to the remaining 20% of advanced security countermeasures. With the continuous evolution of AI capabilities, it will become a consensus in the industry to use "AI shield" to resist "AI spear".

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.