The exponential speed increase of modern computer memory and processor architecture is enabled by design innovations that are continually re-evaluated for exploit. Hacking via manipulating computer memory is a proven tactic dating back to the 1990’s but remains a critical inherent vulnerability in modern systems and is often addressed with methods that result in significant performance restrictions. The era of structuring programs around restrictions in physical memory are long past, but the powerful system specs of today’s production systems integrate predictive functions that allow backdoor attacks underlying a broad category of exploits.
The most fundamental activity of any computer is holding data in memory for analysis or interaction with additional input. From the first primitive use of electrified vacuum tubes to hold binary data, the entry and display of information has been the purpose of computer hardware and software. Contemporary hardware systems use a plethora of methods which enable memory to more quickly handle data input or output. Predicting a program’s use of system memory speeds up the processing of data but enables memory exploits and processor backdoors. Research into the exploitation of predictive data-handling processes in computer hardware have led to the announcement of an abundance of possible, albeit technical and difficult, exploits for cloud computing and internet-connected systems.
New exploits are consistently discovered that leverage methods used for speeding data input and output. Modern systems predict the use of system memory to hasten processing activities, but these prediction algorithms can be harnessed by attackers to access otherwise partitioned segments of system resources containing important data. Research by industry and university teams in exploitation of the JavaScript language, one of the most common computer codes in use, has exposed several exploits for harvesting data running in otherwise inaccessible areas of a system. NetSpectre, a particularly complex variety of memory-exploitation attack, was outlined by researchers at Gaz University of Technology in mid-2018 and relies on sending exploitation web traffic to a system without executing any malicious code on the device memory itself. NetSpectre represents a particularly troubling development as it was able to siphon passwords and encryption data for cloud-hosted services, albeit at slow speeds over long periods of time.
Memory exploitation attacks are largely discovered and publicized by computer science researchers and rely on an incredibly intricate manipulation of systems unlikely to affect average users. Their data harvesting proceeds slowly and is likely only to be employed against large networks where gradual theft can achieve a payoff worthy of the investment represented by the formidable coding needed to enable such an attack. University and cybersecurity analysts researching NetSpectre and other advanced cyberattacks assure the general implementation of complex threats like Spectre variants are not yet a reality. It would be premature to cause a panic over massive risk to memory prediction attacks, but the discovery of exploits reliant on a fundamental process inside so many systems is a call for vigilance. Researchers from Northeastern University and IBM released new findings in late 2018 detailing improvements to the still-theoretical attacks they dubbed SplitSpectre, as the exploitation was divided into two components for easier execution. The same team explained existing patches preventing the execution of Spectre attacks would largely prevent the hack but cautioned that while the likelihood of such a compromise in production systems was remote, the continued revision of the underlying theory could lead to eventual implementation in production environments. Patches to prevent Spectre exploits also cause significant slowdown on host machines, sometimes as high as a 14% drop in system performance. As more and more digital infrastructure is hosted on the cloud, vulnerabilities stemming from Spectre represent a huge payoff if successfully implemented against unpatched systems or hit with a zero-day attack before discovered by security researchers.
A prevailing paranoia can seep into discussions of digital security even when major classes of exploits are discovered by benevolent actors and patched before damage occurs. Spectre currently represents a large but distant storm that may yet bank away from our current hosting and hardware structure. With the payoff of zero-day implementations and the resources available to malicious state and criminal actors, very few successful complex attacks can cause immense disorder and cost. Implementing the most current security fixes and following emerging developments about underlying exploits may be small consolation, but it is often the best practice available. The Spectre threat remains, for now, largely an omen.
