At the 2018 World AI Conference in Shanghai, Tencent Cybersecurity Group and the Cybersecurity Innovation Institute co-published their AI Empowers Cyberspace Security: Patterns and Practices report. The work was sponsored by the Shanghai government.
The rapidly increasing number of mobile devices and cloud applications have dramatically expanded both cyberspace and the anonymity and threats therein. Artificial Intelligence is now being maliciously deployed in a wide range of cyberattacks, including automatic vulnerability detection, malware generation, and automated hacks. American cybersecurity solution firm McAfee estimates that to protect itself against such attacks, a typical enterprise will need to hire 24 percent more cybersecurity staff while also building up more proactive and inventive defenses.
The Tencent report separates cybersecurity into three categories: Network, Online Content, and Network Infrastructure Security. It then applies Gartner’s Adaptive Security Architecture (ASA) proposed in 2014, which identifies four security measures: Predict, Prevent, Detect, and Respond. The 3×4 AI security matrix emerges as follows:
A – Network Security
1) Predict – unsupervised training network spots potential threats; expert system, ML and automation are used for risk assessment;
2) Prevent – automation sets up prevention mechanisms; ML can mislead attackers to protect the more valuable digital assets;
3) Detect – ML and expert system monitors traffic, identifies attack patterns and performs real-time unattended network analysis;
4) Respond – the system analyzes and classifies threats, proceeds with automated or manual intervention, provides direction for follow up recovery and auditing.
AI + Network security applications and case studies:
- Viral and malicious code monitoring and defense: SparkCognition, Cylance, Deep Instinct, Invincea
- Network intrusion and defense: Vectra Networks, DarkTrace, Exabeam, CyberX, BluVector, CSAIL x PatternEx
- Identity recognition: Tencent, Alipay
- Junk mail detection: Netsafe, Gmail
- URL-based malicious site identification: Hillstonenet, Google Safebrowsing, DNSBL, PhishTank, Firefox, IE
- Situational security awareness: Tencent, DBAPPSecurity, 360
B – Online Content Security
1) Predict – DL and NLP break down laws and regulations to set up the content protocol; deep learning is used to assess and forecast risks in different scenarios;
2) Prevent – applied DL upgrades the prevention tools;
3) Detect – NLP, image and video analysis tools are used to recognize and parse content; sentient AI is deployed to screen content;
4) Respond – automation is a crucial part for follow up investigation and audit.
Related applications and case studies:
- Public opinion monitoring: Tencent, Zhongkedianji
- Fake news detection: Tencent, Facebook
- Bad information detection: Tencent, Weibo, Alibaba, Baidu, Facebook, Google
- Fraudulent information detection: Tencent, China Telecom
- Financial risk control: Alipay AlphaRisk
C – Network Infrastructure Security
Due to the complexity of network infrastructure scenarios and interlacing technologies, the report gives an example in smart traffic management.
1) Predict – the system learns through historical data and DL to forecast traffic flow;
2) Prevent – the system performs traffic control and prevention through an optimized layout of traffic facilities; the expert system is deployed in road systems;
3) Detect – CV analysis conducts real-time monitoring of traffic light and vehicles to help reduce traffic jams and dispatch resources needed in case of accidents;
4) Respond – automated systems dig up relevant information and provide guidance for follow-ups.
Related applications and case studies:
- Smart grid construction: Alstom
- Intelligent traffic regulation: Tencent, Baidu
- Urban security monitoring Tencent
The past few years have witnessed substantial breakthroughs in AI empowered cybersecurity. MarketResearch estimates the total AI empowered cybersecurity market will exceed US$35 billion by 2024, with anticipated annual compound growth rate reaching 31 percent from 2017-2024. Irish Research and Markets sees machine learning as the core technology that will gobble up US$6 billion of the market pie.
Source: Synced China
Localization: Meghan Han| Editor: Michael Sarazen