27 March 2021

Analysis Academics, AI, and APTs How Six Advanced Persistent Threat-Connected Chinese Universities are Advancing AI Research

Dakota Cary

Six Chinese universities have relationships with Advanced Persistent Threat (APT) hacking teams. Their activities range from recruitment to running cyber operations. These partnerships, themselves a case study in military-civil fusion, allow state-sponsored hackers to quickly move research from the lab to the field. This report examines these universities’ relationships with known APTs and analyzes the schools’ AI/ML research that may translate to future operational capabilities.Download Full Report

Turning cutting-edge research into operational capabilities is the currency of cyber operations. The vulnerability no one else knows about, one found by someone with highly specific knowledge of a program or coding language, opens a backdoor into an adversary’s most sensitive vault. A better understanding of one technology or technique can give cyber operators an advantage over opponents. Governments benefit from compressing the timelines from discovery to exploitation, more rapidly using the insights of researchers for operations.

Artificial intelligence (AI) and its current dominant paradigm, machine learning (ML), almost certainly will not fundamentally alter competition in cyberspace. That said, AI systems will provide both new terrain for cyber operations—as targets that can themselves be hacked—and new tools of cyber operations, as ML aids offensive and defensive efforts. China’s military-civil fusion strategy takes a holistic approach to development and aims to seamlessly incorporate private resources and developments for state use, with the goal of shortening the pathway for non-governmental research on AI and cybersecurity to strengthen and diversify government operational capabilities.1

There is notable precedent for using university developments in state-sponsored hacking operations. Over the past decade, China’s security services have repeatedly turned to select university faculty to conduct research on cyber techniques and, in some cases, run cyber operations. Collaboration between university faculty and cyber operators illustrates China’s approach to military-civil fusion. This report identifies six universities that previously worked with China’s state-sponsored hacking teams and are now conducting research on the use of ML for cyber capabilities; two universities also host research programs on cyber attack and cyber defense of AI systems. This report summarizes the extent to which these universities, with ties to known state-sponsored hacking teams, might aid China’s efforts in these areas.

Key Findings

Multiple universities with connections to Advanced Persistent Threat (APT) cyber actors are conducting research on the intersection of cybersecurity and ML. All six universities employ faculty who are actively conducting research on ML and cybersecurity. At least one known state-backed hacker is researching how to use ML for anomaly detection, a defensive cybersecurity technique. Shanghai Jiao Tong University’s School of Information Security Engineering, a school with ties to the PLA, hosts a research institute conducting research on offensive and defensive cybersecurity techniques; the director of that institute published an analysis of ML and cybersecurity trends in a Ministry of State Security periodical. Research on cybersecurity and AI is moving from academic journals to strategy forums for China’s security services.

Research conducted on the application of machine learning and AI to cybersecurity is extensive, particularly in the areas of anomaly detection systems, malware classification, behavior analysis, and active defense. Most papers published by faculty from these universities examined how to use machine learning for defensive purposes. Offensively oriented papers noted ML’s application to vulnerability discovery and exploitation—a dual-use technique that can secure or wreck software. One paper published by an author at Xidian University and funded by the Key State Laboratory for Information Security’s Unclassified Projects Fund concluded that ML could bolster cyber defenses and improve vulnerability discovery.

Research on the attack and defense of AI systems was less pervasive than research on applying the technology to cybersecurity. Two schools favored for recruiting cyber operators for specific hacking groups are researching the vulnerability of AI systems. Zhejiang University offers classes on the attack and defense of AI systems, alongside classes on how to write intelligence reports.2 Harbin Institute of Technology is conducting research on the topic, but does not publish its progress.

Governments’ use of new technologies affects their relative power and influence in the modern world. Nations that innovate faster and more effectively often build and sustain an advantage over their rivals. In such a competition, ML has the potential to be a game-changing technology, and both China and the United States are racing to exploit its power. More narrowly, the application of ML techniques to traditional cyber operations may prove to be transformative, altering operations’ practice and amplifying their potency. Since these operations are a fundamental part of modern statecraft, having an operational advantage, even if only for a moment, can yield lasting gains for governments. By examining the research at select universities, analysts and decisionmakers can better determine how China may try to apply AI and machine learning techniques to cyber operations in search of this advantage. If the cross-pollination from academic research teams to fielded operations occurs for ML-enabled cyber capabilities as it did in earlier cyber operations, then understanding the depth and breadth of the schools’ work can shed light on future operational developments and their potential security impacts.

Download Full ReportAcademics, AI, and APTs

No comments: