Why would it be risky to let Huawei build 5G networks? [closed]

Question

I'm wondering why there is so much debate about letting Huawei build 5G networks in Europe and other countries. Shouldn't encryption which is used more and more on the web today make all communication unreadable? Are people just worried about metadata? Thanks.

Answer 1

There is a political and economic question here: Who should be granted the power (via economic might or monopoly, via government contracts for building infrastructure, via government endorsement of standards that prefer the technology already in production by one company but not another, etc.) to determine the technology used by the ubiquitous communications infrastructure of telephony? That question involves issues of consent of the governed, international diplomacy, social rules for power plays, economic theory around monopoly and concentration of power, and so on, and as such is off-topic for crypto.se or security.se, but maybe not for politics.se.

There is a security question here: What powers does the party in control of the infrastructure has to wreak havoc if their handlers should so request? If, hypothetically, under pressure by the NSA, Juniper put a remote management back door into all of their firewalls, and if the Chinese power grid's industrial control network or the Chinese agricultural control network were built out of Juniper routers, would that give the United States of Armed Forces a strategic advantage in warfare against China? That question involves issues of coercion by threat of economic devastation and famine, international diplomacy, and so on, and as such is off-topic for crypto.se, but maybe not for security.se or politics.se.

Finally, there is also a cryptography-related question here: What is the problem with a malicious adversary controlling all the world's networking infrastructure, if all the world's cryptography is designed under the premise of a malicious MITM with the power to intercept, eavesdrop, and forges all the world's traffic?

There are two broad categories of problems that could threaten most applications of crypotgraphy:

• A global passive adversary (GPA), who only eavesdrops and never interferes with traffic, enables efficient collection of mass surveillance of metadata in interpersonal communication like Signal. Most cryptography is designed to conceal the content of conversations, but not the event that conversations happened or when or between whom—as you note, metadata is much harder to conceal. And metadata collection is invisible to the outside: operationally, the network is still allowing communication to happen.

  There are anonymity networks like Tor that aim to conceal metadata with the help of cryptography and a diverse network of operators, but a global passive adversary would also threaten anonymity in low-latency networks like Tor, if the government in control of the GPA is in your threat model. Of course, we're talking about 5G, not the entire internet, but this is already a problem with the Five Eyes network, which is why you should run Tor nodes in as many different jurisdictions and on as many different networks as you can to diffuse the power to monitor the Tor network.

  (Doing this is much easier than persuading everyone to browse the web through high-latency mixnets by email like Richard Stallman does. And the FVEY or hypothetical Huawei threat doesn't mean Tor is useless: there are many threat models other than the US or Chinese governments against which Tor is useful, like abusive spouses, corporate competitors, puritanical parents, etc.)

• A global active adversary like this may have the power to unilaterally deny service to whatever parts of the network they don't like. See this question for some examples of how availability can be related to cryptography. Cryptography can detect forgeries or other active attacks, but it can't prevent them.

  There are many past examples of internet censorship all over the world. Most of them are limited in scope essentially to a single geographical region, like the Great Firewall of China, or like when Mubarak cut off the internet in Egypt in a failed attempt to persuade the people how responsibly he wielded power, or like when the United States Senate discusses an internet kill switch for the sake of freedom—except in cases where like when the Pakistani government decided to block YouTube in Pakistan but accidentally the whole world, but a large-scale block that was quickly reversed.

  Governments are getting better at making censorship unobtrusive, like asking Baidu to quietly discard politically inconvenient search results for Tianenman Square but otherwise function as desired for the vast majority of queries. This happens above the level of mere networking gear, but what if you had the power to just make some things unreliable for parts of the world, like making Vkontakte slower than Facebook—or cut off access to a power substation's 5G-connected remote control system at a convenient moment? Nothing in cryptography can detect these latent powers, nor prevent the consequences of exercising it.

• And, of course, cryptography engineering for end-to-end authenticated encryption in applications takes work and a lot of systems just don't have that work put into them because the economic incentives aren't there. We've made enormous strides in getting billions of people to use more or less end-to-end authenticated encryption for personal messaging in systems like WhatsApp and Signal. What about other applications, like banking, like document management for corporate secrets, like personnel files?

  At resourceful companies where security engineers have clout like Google the premise of a secure network may have been discarded in favor of pervasive end-to-end authenticated encryption, but there is no shortage of corporate environments today that rely on a corporate VPN for access control. If Cisco, or Huawei, or whoever it is, gets essentially exclusive access to the market for 5G, it would be awfully convenient for these corporate networks to be built out of a nice integrated product with 5G and VPN support—and it would be an awfully convenient vector for nationalist industrial espionage or sabotage.