As I traversed the globe for four years as the Cylance Ambassador-at-Large and would share with audiences and customers the prowess that AI-powered machine learning was bringing to the battlespace, I was frequently asked what we could anticipate in the way of a reaction from our adversaries. In the shadow of the SolarWinds compromise, my thoughts reverted to those questions. My response at the time was that we should definitely anticipate a retaliation, pivot or adjustment. There was too much at stake, financially and otherwise, for them not to respond.
My initial impression based on past actions I have seen nation-state adversaries take in response to a perceived advance made in our cyber defenses, was to retreat into the physical world and to advance interests through those vectors. They looked to and exploited “trusted insiders.” In one instance, adversaries reached out and tasked a member of my own security team, charging them to tell no one, particularly their U.S. masters, and to advance the task exclusively in the physical world where indications of compromise were less likely to be detected with the ever-growing sophistication of AI-supported cyber tools.
That I learned of the tasking casts a glimmer of hope on the cynicism that today so often impugns the wisdom of having extended hard-earned trust to certain humans. Although, as speculation as to how SolarWinds was perpetrated matures, the exploitation of a trusted insider must remain in the mix of explanatory theories. Other elements have also persisted: Long and patient reconnaissance; spear-phishing; the exploitation of Continuous Introduction/Continuous Delivery [CI/CD] architectures and efficiency plays; the porosity and extended nature of our modern supply chains [no shortage of weak-links through which to advance adversarial efforts]; Crimeware-as-a-Service and the attribution challenge made even more difficult with ever-increasing conduits of plausible deniability.
Not unlike the compromise I experienced in which an adversary was discovered to have made successful entry two years earlier and then exhibited the prowess and patience to shut down and wait for what they deemed was the optimum moment to strike in advance of sensitive negotiations, the SolarWinds affair was most likely preceded by months if not years of conscientious reconnoitering. Attackers most likely surveyed the software supply chain of companies years in advance. URL endpoints by which they might deliver malware were set up concomitantly. Once SolarWinds was targeted, that patience was punctuated by waiting until their market penetration grew to a huge footprint in the network monitoring domain and amongst government customers.
While exploitation of a trusted insider might well have been leveraged, several other possible attack vectors could have been advanced concurrently. The recent spate of well-crafted, targeted spear-phishing instances could have served as one of those vectors. The sophistication of such efforts has demonstrated a marked uptick in recent years. From there, a successful credential compromise would not have been long in following with its associated internal entry and lateral movement with all that entailed.
Evolving in an equally precarious manner and adding to the complexity evidenced in this compromise, has been the growth and acceptance over the years as the backbone of modern-day DevOps operations — CI/CD, an approach to software development that seeks to leverage shorter development cycles in delivering a steady stream of potentially disruptive innovations to customers who incessantly clamor for “more…faster.” We now grasp the implications of a foundational system, whose updates were compromised and propagated in that manner. The contextual battlespace in which that propagation occurred was further exacerbated by the growing porosity that makes up the modern supply chain — think Boeing 787 — giving an adversary an almost an unlimited number of “weakest leaks” through which to explore the options and realize the fruits of their efforts.
Finally, the emergence, sophistication and plausible deniability associated with Crimeware-as-a-Service now available in the Dark Web and documented by the recent Bahamut Report, means that nation-states can mask their efforts behind third-party contractors and an almost impenetrable wall of plausible deniability. Such actors can obfuscate their efforts in a manner that makes it appear as though they originated practically anywhere, making any quick declarations of attribution — impugning the usual suspects — understandably suspect. At the end of the day, as users of software, companies have witnessed another reason for applying zero-trust networking principles and role-based access controls, not just to users, but also to applications and servers.