On October 4, 2018, Bloomberg published a story titled “The Big Hack: How China Used a Tiny Chip to Infiltrate U.S. Companies,” with a subtitle “The attack by Chinese spies reached almost 30 U.S. companies, including Amazon and Apple, by compromising America’s technology supply chain, according to extensive interviews with government and corporate sources.” From the article:
Since the implants were small, the amount of code they contained was small as well. But they were capable of doing two very important things: telling the device to communicate with one of several anonymous computers elsewhere on the internet that were loaded with more complex code; and preparing the device’s operating system to accept this new code. The illicit chips could do all this because they were connected to the baseboard management controller, a kind of superchip that administrators use to remotely log in to problematic servers, giving them access to the most sensitive code even on machines that have crashed or are turned off.
Companies mentioned in the story deny the details, so this post does not debate the merit of the Bloomberg reporters’ claims. Rather, I prefer to discuss how a computer incident response team (CIRT) and a chief information security officer (CISO) should handle such a possibility. What should be done when hardware-level attacks enabling remote access via the network are possible?
This is not a new question. I have addressed the architecture and practices needed to mitigate this attack model in previous writings. This scenario is a driving force behind my recommendation for network security monitoring (NSM) for any organization running a network, of any kind. This does not mean endpoint-centric security, or other security models, should be abandoned. Rather, my argument shows why NSM offers unique benefits when facing hardware supply chain attacks.
The problem is one of trust and detectability. The problem here is that one loses trust in the integrity of a computing platform when one suspects a compromised hardware environment. One way to validate whether a computing platform is trustworthy is to monitor outside of it, at places where the hardware cannot know it is being monitored, and cannot interfere with that monitoring. Software installed on the hardware is by definition untrustworthy because the hardware backdoor may have the capability to obscure or degrade the visibility and control provided by an endpoint agent.
Network security monitoring applied outside the hardware platform does not suffer this limitation, if certain safeguards are implemented. NSM suffers limitations unique to its deployment, of course, and they will be outlined shortly. By watching traffic to and from a suspected computing platform, CIRTs have a chance to identify suspicious and malicious activity, such as contact with remote command and control (C2) infrastructure. NSM data on this C2 activity can be collected and stored in many forms, such as any of the seven NSM data types: 1) full content; 2) extracted content; 3) session data; 4) transaction data; 5) statistical data; 6) metadata; and 7) alert data.
Most likely session and transaction data would have been most useful for the case at hand. Once intelligence agencies identified that command and control infrastructure used by the alleged Chinese agents in this example, they could provide that information to the CIRT, who could then query historical NSM data for connectivity between enterprise assets and C2 servers. The results of those queries would help determine if and when an enterprise was victimized by compromised hardware.
The limitations of this approach are worth noting. First, if the intruders never activated their backdoors, then there would be no evidence of communications with C2 servers. Hardware inspection would be the main way to deal with this problem. Second, the intruders may leverage popular Internet services for their C2. Historical examples include command and control via Twitter, domain fronting via Google or other Web sites, and other covert channels. Depending on the nature of the communication, it would be difficult, though not impossible, to deal with this situation, mainly through careful analysis. Third, traditional network-centric monitoring would be challenging if the intruders employed an out-of-band C2 channel, such as a cellular or radio network. This has been seen in the wild but does not appear to be the case in this incident. Technical countermeasures, whereby rooms are swept for unauthorized signals, would have to be employed. Fourth, it’s possible, albeit unlikely, that NSM sensors tasked with watching for suspicious and malicious activity are themselves hosted on compromised hardware, making their reporting also untrustworthy.
The remedy for the last instance is easier than that for the previous three. Proper architecture and deployment can radically improve the trust one can place in NSM sensors. First, the sensors should not be able to connect to arbitrary systems on the Internet. The most security conscious administrators apply patches and modifications using direct access to trusted local sources, and do not allow access for any reason other than data retrieval and system maintenance. In other words, no one browses Web sites or checks their email from NSM sensors! Second, this moratorium on arbitrary connections should be enforced by firewalls outside the NSM sensors, and any connection attempts that violate the firewall policy should generate a high-priority alert. It is again theoretically possible for an extremely advanced intruder to circumvent these controls, but this approach increases the likelihood of an adversary tripping a wire at some point, revealing his or her presence.
The bottom line is that NSM must be a part of the detection and response strategy for any organization that runs a network. Collecting and analyzing the core NSM data types, in concert with host-based security, integration with third party intelligence, and infrastructure logging, provides the best chance for CIRTs to detect and respond to the sorts of adversaries who escalate their activities to the level of hardware hacking via the supply chain. Whether or not the Bloomberg story is true, the investment in NSM merits the peace of mind a CISO will enjoy when his or her CIRT is equipped with robust network visibility.