How to Protect Data Center from Cyberattacks – Hidden Costs of Detection-Based Approach for Cloud, Virtualized and Edge Infrastructure

Data centers are truly the life blood of many organizations – whether on-premises or in the cloud –providing access to information and applications that characterise that enterprise. As data centers have become more central and valuable, so too has the job of securing them from cyberattacks. To date, many data centers have followed the familiar path of network security, honing to the perimeter. Unfortunately, this approach leaves serious security gaps, as evident by the proliferation of successful attacks against data center environments.

How to Protect Data Center from Cyberattacks - Hidden Costs of Detection-Based Approach for Cloud, Virtualized and Edge Infrastructure

In today’s heightened cyber threat landscape, the traditional data center security model of perimeter controls and detection-based models does not prevent sophisticated attacks on virtualized IT infrastructure or the OT firmware and software that supports it. That’s because threat prevention technologies are often focused on detecting an initial compromise rather than stopping an attack. They are also heavily reliant on resource-intensive monitoring, which does little to truly reduce risk.

Read this article for an in-depth look at a new approach to traditional data center cybersecurity, known as cyberhardening which shrinks attack surfaces, denies malware the uniformity to propagate, and provides greater protection than perimeter controls and detection-based models. By hardening software binaries, data center security teams can eliminate an entire class of cyberattacks.

After reading this article, those responsible for the integrity, confidentiality and availability of data centers will be well-informed about cyberhardening and how the technique provides greater protection than perimeter controls and detection-based models.

Specifically, this article addresses:

  • The scope, costs and causes of vulnerabilities: how virtualization, smart connected devices, and an increasingly complex supply chain weaken defenses;
  • The mismatch of perimeter/detection security and newer exploit types;
  • Weaknesses in operational technology and information technology infrastructures;
  • How to protect your data center from the inside out by “cyberhardening” software; and
  • The benefits of RunSafe Security’s Alkemist, which uses Runtime Application Self-Protection to prevent exploits from executing by hardening and randomizing software binaries.

Content Summary

The scope and cost of vulnerabilities in data centers
Causes of the increase in data center vulnerabilities that enable cybercrime
The problem with perimeter security and a detection-based approach
Managing security beyond detection in data center environments
Forget the perimeter, bake in security to harden data center defenses
RunSafe Security’s Alkemist protects software from the inside with RASP
The Alkemist transformation process
Protecting your organization with Alkemist

The scope and cost of vulnerabilities in data centers

$260,000 – that’s what Information Technology Intelligence Consulting reports that an hour of downtime costs.

A more sobering statistic is the cost of a data center breach, which includes detection and escalation, notification, post data breach response, and lost business. According to the 2018 Cost of a Data Breach Study (benchmark research sponsored by IBM Security, conducted by the Ponemon Institute), the total average cost in the U.S. was $7.91 million for a single breach, with 25,000+ records compromised. Further, the mean time to identify an attack was an astounding 201 days, with another 52 days to contain. That translates to about seven months during which a bad actor could be determining how to exfiltrate data (and doing so), and another seven weeks to fix the issue, once it has been flagged.

While the value of personal information being sold on the dark web varies depending on the how comprehensive the record is, an estimate of $10 – $150 is reasonable. Assuming an average of $80 per record, even the most modest data centers contain a gold mine for hackers, which serves as ample motivation to “break in.”

Two of the most widely known data breach examples are Marriott and Equifax. In November 2018, Marriott International announced that hackers had stolen data on about 500 million customers. It was later learned that the breach actually happened starting in 2014 in systems supporting Starwood Hotel brands. Some combination of contact information, passport numbers, and credit card data were harvested. Estimates of the financial impact to Marriott range from $2.1 billion to $3.5 billion. In September 2017, Equifax disclosed that an application vulnerability on one of its websites led to a data breach that exposed information about 147.9 million consumers. Data gathered included Social Security numbers, birth dates, addresses, and driver’s license numbers. The financial impact to Equifax was estimated at $90 million.

Causes of the increase in data center vulnerabilities that enable cybercrime

The unintended consequence of cloud-based infrastructure, virtual environments, and edge computing is the rapid expansion of the attack surface, propagated by the software that is necessary to virtualize machines. As such, attackers continually refine their techniques to take advantage of the millions of identical binary templates for virtual environments (a.k.a., golden images) that enable the benefits that so many businesses and government agencies rely on. CSO by IDG noted that virtualization software bugs skyrocketed 275 percent in 2018 over the previous year.

Furthermore, billions of connected devices – including building management systems, thermostats, and all kinds of sensors – now comprise the Internet of Things (IoT). Such mainstream connectivity has facilitated the greatest expansion of the attack surface that the world has ever seen, introducing unprecedented threats of Distributed Denial of Service (DDoS), buffer-overflow, memory corruption and zero-day attacks against industrial, commercial, medical, military and consumer systems and devices. In fact, 70 percent of the most commonly used IoT devices contain vulnerabilities (HP).

Thirdly, the complexity and global nature of the supply chain is now widely regarded as the greatest emerging threat to industry. Every point in a supply chain presents a potential weakness for cybersecurity. And every person that comes in to contact with each piece of hardware or software is a potential threat. 56 percent all breached organizations cite supply chain vulnerabilities as the precursor to the exploits, malware and zero days executed to steal lucrative information, intellectual property, trade secrets and more. And supply chain attacks surged 200 percent in 2017.

None of these risks are likely to diminish anytime soon. Sophisticated attackers understand that globalization continues to expand opportunities for pragmatic supply chain partnerships. Concurrently, inefficient standards and regulations vary, or simply don’t exist, from industry to industry and country to country, leaving an unprecedented number of vulnerable third-parties to exploit.

The problem with perimeter security and a detection-based approach

The perimeter-based model of network and endpoint detection has a problem. Newer types of cyberattacks (zero-days) do not share the security industry’s historical preoccupation with networks. Instead, zero-days target weak links in software directly. The recent Not Petya and WannaCry attacks, for example, were not detected or prevented by traditional cybersecurity controls. These attacks were able to scale rapidly, take down services, and drive up costs for government and commercial organizations around the world. Expenses and delays came from the attacks themselves along with detection-related alerting, triaging, after action reporting, and out-of-band patching.

The very success of firewalls, intrusion prevention, endpoint protection, encryption, and signing solutions have driven adversaries to evolve their techniques to defeat them. Presently, newer attack types leverage vulnerabilities in global supply chains and the vulnerabilities of software and firmware to memory-based (file-less) attacks. File-less attacks, which accounted for 77 percent of all compromised attacks in 2017, take advantage of existing vulnerabilities rather than installing malicious software. Such attacks are almost ten times more likely to succeed than traditional attacks, such as through an executable program copied onto the target’s machine, often disguised as a different file format such as a PDF or JPEG, or hidden inside a carrying file like a compressed zip file.

The security industry’s response to evolving attacker techniques has been to bolt on even more detection technology into networks or hosts, leading to even more alerts and false positives. This approach, while good in theory, fails to take into consideration whether or not security can be incorporated into the software itself as a means to prevent the root cause of attacks. If possible, the endless cycle of new and successful attacks and costly treatment of symptoms after the fact could finally be broken.

Managing security beyond detection in data center environments

Today’s on-premises and cloud data centers represent a multi-vendor model sitting at the end of long global supply chains. Every day brings more updates, and with them potentially more vulnerabilities. Let us look at the data center from two points of view – firstly the foundational layer of Operation Technology (OT) hardware and secondly the software that makes up the Information Technology (IT) stack on top.

Figure 1 shows typical OT hardware whose software and firmware underpin the data center.
Figure 1 shows typical OT hardware whose software and firmware underpin the data center.

OT provides the IT infrastructure with everything it needs from power, heating, and cooling to its bare metal foundations. Each OT vendor in turn has dozens of sub vendors. Vulnerabilities in the OT layer can lead to compromise of even off public network (air-gapped) resources, as happened with Target’s financial systems. The hackers who got away with information on 40 million payment cards accessed Target’s corporate network using malware and a connection through an HVAC contractor. Similarly, a smart device in the lobby aquarium of a North American casino had been remotely monitoring temperature, salinity, and automatic feedings. That internet-connected instrument allowed hackers to exfiltrate 10 gigabytes of high roller data. In addition, a thermostat was found to be the weak link in a Chinese attack on the U.S. Chamber of Commerce.

Figure 2 shows the IT infrastructure built on top of the OT infrastructure.
Figure 2 shows the IT infrastructure built on top of the OT infrastructure.

Hypervisors and virtualization form a Hyperconverged Infrastructure (HCI) to manage enterprise storage, servers, and network resources for maximum efficiency and maintainability. These resources are in turn accessed from a variety of external devices such as smartphones, tablets, laptops, desktops, and IoT devices.

The exact same software “golden images” for enterprise IT and OT infrastructure software are used hundreds of times across production, disaster recover (DR), and Continuity of Operations (COOP) sites. For industry standard commercial and Open Source Software (OSS) components, there may be millions of identical copies around the world. The economies of scale that use standard images are beneficial to not just organizations, but also to adversaries.

A single exploit can be run standalone, or in combination with other techniques (a so-called “kill chain”) to spy, steal financial data, run ransomware, or launch a DDoS attack that scales across organizations and around the world. The modern data center is essentially a software monoculture that lacks natural resistance and diversity since, in theory, it used to sit within a protective perimeter. However, that perimeter has proven porous despite steady improvements of detection with Artificial Intelligence (AI), Machine Learning (ML), and other techniques.

Forget the perimeter, bake in security to harden data center defenses

In the past, baking cybersecurity into software meant more detection, manual, static, and/or dynamic analysis (SAST/DAST), and then re-engineering source code. It was a heavy lift, used in only a few niche areas. In many cases, the effort was not so much prevention as a movement of detection agents from the network to the hosts (NIDS to HIDS). Results were typically limited in value, since covering only a fraction of the source code, missing vulnerabilities hidden within third-party software such as frameworks, middleware, and libraries, or just present in different layers of the infrastructure not covered by specialized web or operating system (OS)-based use cases.

Memory corruption attacks try to trick a software program into running attacker-provided code, instead of programmer-written code. For this to work, the attacker must find vulnerabilities in the software binary code that allows the injection of code and/or the redirection of execution. Runtime Application Self Protection (RASP) is a security technology that uses runtime instrumentation to detect and block cyberattacks by employing information from inside the running software. This differs from perimeter-based protections such as firewalls, that can only detect and block attacks by using network information without contextual awareness. RASP technology improves the security of software by monitoring its inputs, and blocking those that could allow attacks, while protecting the runtime environment from unwanted changes. When a threat is detected, RASP can prevent exploitation and possibly take other actions, including terminating a user’s session, shutting down the application, alerting security personnel and sending a warning to the user. RASP can close the gap left by application security testing and network perimeter controls, neither of which have enough insight into realtime data and event flows to prevent vulnerabilities from slipping through the review process or block new threats that were unforeseen during development.

RunSafe Security’s Alkemist protects software from the inside with RASP

To mitigate the vulnerabilities of perimeter security that leave data centers at significant risk, RunSafe Security hardens software binaries using RASP techniques. The technology process is termed cyberhardening, and it combats memory corruption errors and buffer overflow exploits – the weaknesses that attackers typically use to gain control of embedded systems and devices that are key to today’s data center. Known as Alkemist, our cyberhardening technology is a remotely deployable, self-service transformation process that can be accessed through a web client or RestAPI.

Alkemist utilizes the following Runtime Application Self-Protection (RASP) techniques:

  • Block-level binary stirring (randomization), which makes each protected system and device functionally identical yet logically unique.
  • Control Flow Integrity (CFI), which protects against Return/Jump Oriented Programming (ROP/JOP) attacks, where existing code is called out of order to become a hacking script. This prevents malware from changing how commands are executed.
  • Stack Frame Randomization, which creates stack-level entropy by randomizing the buffer set aside from local variables when functions are instantiated on the stack. This entropy deprives an attacker of the information needed to craft a payload that weaponizes a stack overflow vulnerability.

Thus, every line of code and memory block that is called or utilized has randomization and hardening applied to thwart an attacker from understanding the memory topology of a device for malicious replication across a system. All of these actions are performed once per binary and/or in real time with a patented low overhead footprint. This hardening and stirring results in shrinking the attack surface. There are no changes made to the compiler, OS, application, or variable inputs. As a result, legacy infrastructure systems remain untouched with RunSafe’s Alkemist serving as a shield.

By precluding an exploit from spreading across multiple devices and networks, Alkemist eliminates an entire class of cyberattacks, providing protection against downtime and data exfiltration. Additionally, by hardening binaries (in other words, baking in security), Alkemist denies malware the uniformity necessary to execute and propagate, thereby preserving software developers’ intent.

The Alkemist transformation process

RunSafe’s Alkemist™ transforms software quickly, easily, and directly. It adds hardening and diversity with low overhead, leaving the functionality unchanged.

Figure 3: Alkemist can be used at any layer from IT to OT
Figure 3: Alkemist can be used at any layer from IT to OT

Alkemist’s patented Runtime Application Self-Defense (RASP) techniques are specifically designed to prevent the modern fileless, memory corruption, ROP (return oriented programming), and supply chain attacks that have evolved to bypass detection-based approaches.

Protecting your organization with Alkemist

Alkemist can be accessed from the cloud or on-premises, and can be automatically applied to new builds or systems already fielded. It covers software down to bare metal, leveraging Windows, Linux, or no operating system, across AMD, ARM, Intel, and PowerPC platforms. Most significantly, it does not require access to source code, and leaves functionality unchanged. It is a new layer of protection, complimenting existing defense-in-depth tools and processes.

Cyberattacks causing data center downtime and/or the exfiltration of personal information wreak havoc on organizations. These impacts include business disruption, reputational damage, loss of productivity, and significant negative financial repercussions. Virtualization, the prevalence of smart connected devices, and an increasingly complex supply chain contribute to the expansion of the attack surface. Perimeter defenses which worked previously have encouraged the evolution of software exploits which are no longer mitigated by traditional methods.

Source: RunSafe Security