Dimensionality Reduction Attacks
降维打击 and Chinese Perspectives on High-Tech Competition
During the pandemic I read the Chinese science fiction trilogy “Remembrance of Earth’s Past”, better known for its opening novel “The Three-Body Problem”. Author Liú Cíxīn’s work was brilliant and instantly became my favorite science fiction book.
If you haven’t read any of these novels, buy the books and start tonight. Some of what follows contains important spoilers for the plot.
As part of my cybersecurity work I check about a dozen Chinese sources daily. Earlier this year I was reading an article discussing the woes of the Japanese auto industry. The author explained how Japan’s brands had made internal mistakes, but had also been subjected to an exterior blow: the rapid rise of China’s automobile industry. In a matter of just a few years, Chinese car companies shifted from the low-quality-low-price quadrant to the high-quality-low-price quadrant: a nightmare for companies like Honda or Nissan.
The overnight technological superiority of Chinese brands has been so staggering that the author described it as a “dimensionality reduction attack” (降维打击, jiàngwéi dǎ jī), an expression adopted directly from Liú’s science fiction work.
I warned there would be spoilers. In the novels, a dimension-reduction attack is a shocking, unimaginable, terrifying development. It is conducted by an entity that enjoys an inconceivable technological advantage over the target. There is no warning and, most bitterly, no defense against such an attack.
I wasn’t aware that the 降维打击 concept was used outside of its science fiction origins (for convenience, I will also refer to it as “DRA”). In the context of international auto tech competition it is of course an exaggeration: Japanese industry won’t disappear overnight. However, as a literary device the expression is very effective. The Chinese tech analysts who use the idiom do it to convey that existential-level punches are being exchanged between companies and countries.
Dynamism and high stakes are the norm in high tech competition: companies and products constantly age out as newcomers and innovators shake things up. What’s new in the 2020s is the entry of China’s many labs and companies to the playing field.
Where else has the 降维打击 DRA concept been used? I checked and observed the following examples, all taken from Chinese sources:
Many associated with the auto industry:
BYD’s 2021 launch of DM-i super hybrid technology constituted a DRA against the gasoline-based auto industry
Nvidia’s 2022 announcement of the Thor superchip constituted a DRA against competing automobile CPU/GPU makers
Xiaomi’s 2024 approach for the SU7 constituted a DRA against standard auto marketing practices
Black Sesame Technologies achieved dimensionality reduction against its rivals via its L4 automated driving systems
A few AI ones from this year:
In January 2026, the release of DeepSeek-OCR 2 constituted a DRA against Adobe and other legacy OCR providers
In March 2026, someone advertising the OpenClaw fork WindClaw described it as a DRA against the original, due to its supposedly superior UX
Also in March 2026, Nvidia’s announcement of AI-RAN (shifting 5G/6G base station processing to virtualized cloud-hosted environments) was discussed as a DRA against China’s Huawei
I also found some examples unrelated to high tech:
The 2005 launch of Xiang Piao Piao’s milk tea was a DRA against “many other street milk tea shops”
Chinese fintech applications failed to conduct a DRA against India’s fintech firms due to local complexities and regulations
High-profile artists are described as having dimensionality reduction singing and dancing abilities, or choosing to explore less saturated artistic media where DRAs are more achievable
A few conclusions from my sampling of these various articles:
DRAs are rapid, panic-inducing changes in the balance of technological power between companies or countries
The term mostly describes high-tech competition in semiconductors, EVs, or AI models
It applies less frequently to non-tech activities, likely due to its sci-fi origins
It’s unclear to what extent the expression 降维打击 is immediately understandable to general Chinese audiences
All sampled uses of the expression originated in individual writers rather than organizations, much less the government
Usage may not just be nationalist but also in the interests of a specific company
The term is most often used to illustrate a Chinese company becoming technologically supreme, though not always (as illustrated by the two Nvidia examples)
Lastly, a word on information security. As of 2026 China continues to punch below its weight in cyber. China’s national cybersecurity programs, as formidable as they may be, are still modest relative to the country’s economic and technological prowess, not to mention sheer size. Evidence of fragmentation and even immaturity often emerges, such as military purges, the lack of a national agency, or contractors shopping their breaches around various government agencies.
China exports little in terms of defensive cybersecurity products and technologies. Despite the defensive benefits of the Great Firewall and its Traffic Secure Gateways, reliable reports emerge as recently as this month that China’s top science and technology institutions are regularly breached from the outside. Chinese software, even from prestigious brands, is often insecure by design (real-world example: software conducting OCR on a user’s entire desktop surface and transferring the data in plain HTTP).
On the offensive side, while Chinese APTs have a consistent record of success, so far China has been more on the receiving end of DRA-level network intrusions than on the delivery side. Leaks from recent years suggest internal Communist Party deliberations and other sensitive institutions have been breached. Just last week the US president said that Washington routinely executes significant breaches of Chinese networks. However, there are no resource or technological constraints stopping China from developing an NSA-like network intrusion capability. The country lacks neither talent nor money to pursue one. Rather, it is Beijing’s current policy choices and organizational frameworks that determine how quickly China will upgrade its offensive cyber capabilities.
Turning now to pure speculation, what would Chinese dimensionality-reduction capabilities look like in cyber?
In offensive terms the nightmare scenario is not “Chinese APT leverages phishing and/or malware to breach an important institution” - that’s just linear growth and is already happening today.
A true game changer would involve delivering undetectable malware more directly, via compromised trusted connections, intercepted DNS routing, operating system update mechanisms, or official app stores. In other words, the kind of supply chain attacks that Beijing conducts on its domestic Internet, but at scale and internationally. This is why so many in the West object to Chinese companies having a foothold in critical networking infrastructure. Another offensive DRA would be the equivalent of the Claude Mythos LLM that’s in the news recently: an AI-powered system for discovering, and later exploiting, vulnerabilities in key software at scale.
Dimensionality reduction is an offensive-oriented concept, so it’s harder to think of defensive innovations that could replicate its spirit. When it comes to great power competition, China improving its network defenses to decisively kick the NSA and other foreign adversaries out would come close enough.
Stacking additional defenses is the defensive version of linear growth. In order to truly change the nature of the battle, China might expand its technological independence drive to software (hardware is well underway) and develop local alternatives to everything. If no more western operating systems, backbone and cloud providers, programming languages, or even open-source libraries were used, and instead China operated its own closed-source fleet of technologies, it would impose new challenges on Beijing’s adversaries. There are hints of this taking place, such as aspirational language seen in five-year plans, or anecdotes like last year’s rare earths regulations being published in Kingsoft WPS format rather than .DOC or .PDF. However, a categorical decision to disengage from the Western IT stack has not been made: it will be much more evident if and when it happens.
This is all pure speculation - I’m sure there are different, perhaps more likely scenarios that would come to mind to someone with access to different information or greater creativity.