Dronegate: The First Casualty is Our Cybersecurity Paradigm

As of yet, there is no definitive narrative of the virus that hit the U.S. drone fleet at Creech Air Force Base in Nevada this September. Original reports stated that drone cockpits had been infected with a keylogger virus and, while there was no indication that classified information had been stolen or that missions had been compromised, the virus has proven tenacious, resisting efforts to disinfect machines and forcing the Air Force to wipe entire hard drives. Sources said that officials at Creech never informed the 24th Air Force, the central authority on cyber for the Air Force, about the breach until the 24th read about it online. Yesterday, however, in its first official statement on the infection, the Air Force explained that the virus was actually credential stealer and insisted that the virus was only a nuisance that was easily contained. It claimed that the 24th AF had known about the breach since the 15 September. The Air Force also disputed that cockpits were affected, stating that only ground control systems were breached.

If initial reports were true, then our military cybersecurity is in a lamentable state. The most critical element of perhaps our most vital weapons and intelligence systems would have been breached, and the primary defenders were kept in the dark because of the fear of failure that permeates security and stifles information-sharing and cooperation. But even if the relatively optimistic official accounts of the infection are the whole truth, the military’s computer security paradigm still needs an overhaul.

In some ways, the official statement is more worrying than even the most sensational initial accounts as it suggests a disconnect from cybersecurity realities. First, it’s too quick to dismiss what may have been a real threat. According to Microsoft security architects, once a credential stealer gets a foothold on your network, it typically takes between 24 and 48 hours to gain Domain Admin credentials and access to every account and workstation. An anonymous official has claimed that the malware only targets online gaming accounts, but this has not been confirmed or attributed. If the 24th managed to isolate the virus, they may have squashed a nuisance or they averted a crisis. Their confidence in defensive measures is even more unsettling. “Our tools and processes detect this type of malware as soon as it appears on the system, preventing further reach,” the release claims, “We continue to strengthen our cyber defenses, using the latest anti-virus software and other methods.” That the Air Force feels safe behind a cyber Maginot Line, as Professor Rick Forno would say, does not fill me with confidence, especially when the virus has already penetrated “air gaped” systems, the gold standard in network security.

It’s time the Air Force adopts industry best-practice and switches to a “presumption of breach” mindset. Rather than putting all of its energy into keeping all attackers out with technological silver bullets, the Air Force, like top private firms, must assume that it will be infected and most likely already is. This is hardly a stretch. The official release states that drone systems are not facing any “advanced persistent threat” or even targeted attack, just one of millions of random, run-of-the-mill viruses floating around on the internet. The malware in question is said to be commonly used to steal log-ins and passwords for online games, implying that it was picked up in such a setting. How many other isntances of malware were accidently picked up by Air Force personel and possibly transfered on to classified systems? And if malware designed to steal your Mafia Wars account can access some of the military’s most mission critical systems, how long will it take for a sophisticated, state-sponsored virus like Stuxnet makes in on to UAV infrastructure?

A “plan to fail” approach would shift emphasis to forensics and remediation, areas where the Air Force seems to be lagging. If the 24th AF really did know about the infection since 15 September, at least their monitoring and intrusion detection systems are in order. The official release, however, does not say that they have finished disinfecting computers or that they have determined the source of the malware, implying that they are still working on forensics and infection turnaround a month later. If true, the initial insider reports of persistent and mysterious malware confirm this, and add that the only cure seemed to be to wipe internal hard drives and start clean, a costly and time consuming process. It is also important, when operating under a presumption of breach, to share information about infections, attacks, and mistakes, unlike initial reports suggested, rather than hide possible failures and to learn from them. From the tone of the press release, classifying the attack as a minor annoyance promptly taken care of with the latest and greatest technology, it doesn’t look like much learning is taking place.