3 ms·

I can't confirm it because the descriptions of the hack are unclear but if more network operators say they've been hacked it is more and more likely the Chinese

by Zigurd 2y ago

I can't confirm it because the descriptions of the hack are unclear but if more network operators say they've been hacked it is more and more likely the Chinese got in by attacking lawful intercept. This could happen in various ways: bribe or blackmail someone in law enforcement with access to a lawful intercept management system (LIMS), a supply chain attack on an LIMS vendor, hacking the authentication between networks and LIMS, etc. If it is an LI attack the answer to which networks are compromised is: All of them that support automated LI. That's a nasty attack because LI is designed to not be easily detectable because of worries about network operators knowing who is being tapped.

trod1234 2y ago
That reasoning is dubious. They aren't saying that more have been hacked, they are saying that more have been discovered related to that hack. Any adversary at this level would be monitoring the news, and would take appropriate actions (for gain) or roll up the network rather than allow reverse engineering of IOCs. More than likely this was not an LI based attack, but rather they don't know for sure how they got in. Nearly all of the guidance is standard cybersecurity best practices for monitoring and visibility, and lowering attack surface with few exceptions (in the CISA guidance). The major changes appear to be the requirements to no longer use TFTP, and the referral to the manufacturer for source of truth hashes (which have not necessarily been provided in the past). A firmware based attack for egress/ingress seems very likely. For reference, TFTP servers are what send out the ISP configuration for endpoints in their network, the modems (customers), and that includes firmware images (which have no AAA). Additionally as far as I know the hardware involved lacks an ability to properly audit changes to these devices (by design), and TR-47 is rarely used appropriately, the related encryption is also required by law to be backward compatible with known broken encryption. There was a good conference talk on this a few years ago, at Cyphercon 6. https://www.youtube.com/watch?v=_hk2DsCWGXs https://www.youtube.com/watch?v=_hk2DsCWGXs The particular emphasis on TLS1.3 (while now standard practice) suggests that connections may be being downgraded, and the hardware/firmware at CPE bridge may be performing MITM to public sites in earlier versions transparently, if this is the case (its a common capability needed). The emphasis on using specific DH groups, may point to breaks in the key exchange of groups not known to be broken (but are broken), which may or may not be a factor as well. If the adversary can control, and insert malicious code into traffic on-the-fly targeting sensitive individuals who have access already, they can easily use information that passes through to break into highly sensitive systems. The alternative theory while fringe, is maybe they've come up with a way to break feistel networks (in terms of cryptographic breaks). Awhile back the NSA said they had a breakthrough in cryptography. If that breakthrough was related to attacks on feistel network structures (which almost all modern cryptography is built on), that might explain another way (although this is arguably wild speculation at this point). Nearly every computer has a backdoor co-processor built-in in the form of Trustzone, Management Engine, or AMD's PSP. Its largely only secured by crypto without proper audit trails. It presents a low hanging concentrated fruit into almost every computation platform on earth, and by design, its largely not auditable or visible. Food for thought. Quantum computer breaks a single signing key for said systems, acting like a golden key back door to everything. All the eggs in one basket. Not out of the realm of possibility at the nation state level. No visibility means no perception or ability to react, or isolate the issues except indirectly.
- foobiekr 2y ago
  You don’t need to bring up quantum computers. Almost all protocols in the networking industry are basically running with a shared secret that is service global. Pop any box at all and you have the world for any traffic you can capture. The problem with the shared secret model isn’t that it can be stolen, it’s that it is globally shared within a provider network. You can’t root it in a hardware device. You can’t do forensics to see from what node it was stolen. We are talking about an industry where they still connect console servers, often to serial terminal aggregators that are on the internal network alongside the management Ethernet ports, which have dumb guessable passwords, often the same one on every box, that all their bottom tier overseas contractors know. It’s just sad.
  - trod1234 2y ago
    > You don't need to bring up quantum computers. Almost all protocols in the networking ... Its true that those protocols are basically running shared secrets, but those areas all have some visibility with auditing and monitoring. You crack a root or signing key at the co-processor level and you can effectively warp and control what anyone sees or does with almost no forensics being possible. It fundamentally allows a malevolent entity the ability to alter what you see on the fly with no defense possible. Such is the problem with embedded vulnerabilities, its just like that NewAg train thing. Antitrust and bricking for monopolistic benefit is far more newsworthy then say embedding a remote radio-controlled off-switch with no plausible cover, that can brick the trains as they move harvests, food stuffs, or military equipment. Its corruption, not national security. Would many believe that its the latter over the former when it does both? It is sad that our societal systems have become so brittle that they cannot check or effectively stop the structural defects and destructive influences within itself.
foobiekr 2y ago
More likely they got access and then snooped any of the many insecure protocols used to manage network devices. Anyone who has ever worked in networking will understand what I mean. The networking industry is comically bad. They use ssh but never ever verify host keys, use agent forwarding, use protocols like RADIUS or SNMP which are completely insecure once you pop a single box and use the almost always global shared secret. Likewise the other protocols. Do they use secure boot in a meaningful way? So they verify the file system? I have news for you if you think yes. It’s kind of a joke how bad the situation is. Twenty years ago someone discovered you could inject forged tcp resets to blow up BGP connections. What did the network industry do? Did they institute BGP over TLS? They did not. Instead they added TCP MD5 hashing (rfc: https://datatracker.ietf.org/doc/html/rfc2385 https://datatracker.ietf.org/doc/html/rfc2385 in 1999) using a shared secret because no one in networking could dream of using PKI. Still true today. If deployed at all, which it usually isn’t. 2010!! If you want to understand the networking industry consider only this: instead of acknowledging how dumb the situation is and just using tls, instead we got this - https://datatracker.ietf.org/doc/html/rfc5925 https://datatracker.ietf.org/doc/html/rfc5925 - which is almost as dumb as 2385 and just as bad in actual deployment because they just keep using the same deployment model (the shared tuple). Not all vendors that “support” 5925 support the whole RFC. As an aside this situation is well known. People have talked about it for literal decades. The vendors have shown little to no interest in making security better except point fixes for the kind of dumb shit they get caught on. Very few security researchers look at networking gear or only look at low end junk that doesn’t really matter.
- immibis 2y ago
  FWIW PKI tends to mean a central point of failure. Some Russian organizations can't get TLS certificates because of sanctions.
  - foobiekr 2y ago
    Pki here does not mean a global CA. You can run your own CAs (and should).
    - 2y ago

by Zigurd 2y ago

trod1234 2y ago

That reasoning is dubious. They aren't saying that more have been hacked, they are saying that more have been discovered related to that hack. Any adversary at this level would be monitoring the news, and would take appropriate actions (for gain) or roll up the network rather than allow reverse engineering of IOCs. More than likely this was not an LI based attack, but rather they don't know for sure how they got in. Nearly all of the guidance is standard cybersecurity best practices for monitoring and visibility, and lowering attack surface with few exceptions (in the CISA guidance). The major changes appear to be the requirements to no longer use TFTP, and the referral to the manufacturer for source of truth hashes (which have not necessarily been provided in the past). A firmware based attack for egress/ingress seems very likely. For reference, TFTP servers are what send out the ISP configuration for endpoints in their network, the modems (customers), and that includes firmware images (which have no AAA). Additionally as far as I know the hardware involved lacks an ability to properly audit changes to these devices (by design), and TR-47 is rarely used appropriately, the related encryption is also required by law to be backward compatible with known broken encryption. There was a good conference talk on this a few years ago, at Cyphercon 6. https://www.youtube.com/watch?v=_hk2DsCWGXs https://www.youtube.com/watch?v=_hk2DsCWGXs The particular emphasis on TLS1.3 (while now standard practice) suggests that connections may be being downgraded, and the hardware/firmware at CPE bridge may be performing MITM to public sites in earlier versions transparently, if this is the case (its a common capability needed). The emphasis on using specific DH groups, may point to breaks in the key exchange of groups not known to be broken (but are broken), which may or may not be a factor as well. If the adversary can control, and insert malicious code into traffic on-the-fly targeting sensitive individuals who have access already, they can easily use information that passes through to break into highly sensitive systems. The alternative theory while fringe, is maybe they've come up with a way to break feistel networks (in terms of cryptographic breaks). Awhile back the NSA said they had a breakthrough in cryptography. If that breakthrough was related to attacks on feistel network structures (which almost all modern cryptography is built on), that might explain another way (although this is arguably wild speculation at this point). Nearly every computer has a backdoor co-processor built-in in the form of Trustzone, Management Engine, or AMD's PSP. Its largely only secured by crypto without proper audit trails. It presents a low hanging concentrated fruit into almost every computation platform on earth, and by design, its largely not auditable or visible. Food for thought. Quantum computer breaks a single signing key for said systems, acting like a golden key back door to everything. All the eggs in one basket. Not out of the realm of possibility at the nation state level. No visibility means no perception or ability to react, or isolate the issues except indirectly.

foobiekr 2y ago
You don’t need to bring up quantum computers. Almost all protocols in the networking industry are basically running with a shared secret that is service global. Pop any box at all and you have the world for any traffic you can capture. The problem with the shared secret model isn’t that it can be stolen, it’s that it is globally shared within a provider network. You can’t root it in a hardware device. You can’t do forensics to see from what node it was stolen. We are talking about an industry where they still connect console servers, often to serial terminal aggregators that are on the internal network alongside the management Ethernet ports, which have dumb guessable passwords, often the same one on every box, that all their bottom tier overseas contractors know. It’s just sad.
- trod1234 2y ago
  > You don't need to bring up quantum computers. Almost all protocols in the networking ... Its true that those protocols are basically running shared secrets, but those areas all have some visibility with auditing and monitoring. You crack a root or signing key at the co-processor level and you can effectively warp and control what anyone sees or does with almost no forensics being possible. It fundamentally allows a malevolent entity the ability to alter what you see on the fly with no defense possible. Such is the problem with embedded vulnerabilities, its just like that NewAg train thing. Antitrust and bricking for monopolistic benefit is far more newsworthy then say embedding a remote radio-controlled off-switch with no plausible cover, that can brick the trains as they move harvests, food stuffs, or military equipment. Its corruption, not national security. Would many believe that its the latter over the former when it does both? It is sad that our societal systems have become so brittle that they cannot check or effectively stop the structural defects and destructive influences within itself.

foobiekr 2y ago

More likely they got access and then snooped any of the many insecure protocols used to manage network devices. Anyone who has ever worked in networking will understand what I mean. The networking industry is comically bad. They use ssh but never ever verify host keys, use agent forwarding, use protocols like RADIUS or SNMP which are completely insecure once you pop a single box and use the almost always global shared secret. Likewise the other protocols. Do they use secure boot in a meaningful way? So they verify the file system? I have news for you if you think yes. It’s kind of a joke how bad the situation is. Twenty years ago someone discovered you could inject forged tcp resets to blow up BGP connections. What did the network industry do? Did they institute BGP over TLS? They did not. Instead they added TCP MD5 hashing (rfc: https://datatracker.ietf.org/doc/html/rfc2385 https://datatracker.ietf.org/doc/html/rfc2385 in 1999) using a shared secret because no one in networking could dream of using PKI. Still true today. If deployed at all, which it usually isn’t. 2010!! If you want to understand the networking industry consider only this: instead of acknowledging how dumb the situation is and just using tls, instead we got this - https://datatracker.ietf.org/doc/html/rfc5925 https://datatracker.ietf.org/doc/html/rfc5925 - which is almost as dumb as 2385 and just as bad in actual deployment because they just keep using the same deployment model (the shared tuple). Not all vendors that “support” 5925 support the whole RFC. As an aside this situation is well known. People have talked about it for literal decades. The vendors have shown little to no interest in making security better except point fixes for the kind of dumb shit they get caught on. Very few security researchers look at networking gear or only look at low end junk that doesn’t really matter.

immibis 2y ago
FWIW PKI tends to mean a central point of failure. Some Russian organizations can't get TLS certificates because of sanctions.
- foobiekr 2y ago
  Pki here does not mean a global CA. You can run your own CAs (and should).
  - 2y ago