Posts

RTDMI Evolving with Machine Learning to Stop ‘Never-Before-Seen’ Cyberattacks

/1 Comment/in , /by

If I asked you, “How many new forms of malware did SonicWall discover last year?” What would be your response?

When I pose this question to audiences around the world, the most common guess is 8,000. People are often shocked when they hear that SonicWall discovered 45 million new malware variants in 2018, as reported in the 2019 SonicWall Cyber Threat Report.

The SonicWall Capture Labs threat research team was established in the mid-‘90s to catalog and build defenses for the massive volume of malware they would find each year. Because our threat researchers process more than 100,000 malware samples a day, they have to work smart, not hard. This is why SonicWall Capture Labs developed technology using machine learning to discover and identify new malware. And it continues to evolve each day.

How Automation, Machine Learning Stops New Malware

Released to the public in 2016, the SonicWall Capture Advanced Threat Protection (ATP) sandbox service was designed to mitigate millions of new forms of malware that attempt to circumvent traditional network defenses via evasion tactics. It was built as a multi-engine architecture in order to present the malicious code different environments to detonate within. In 2018, this technology found nearly 400,000 brand new forms of malware, much of which came from customer submissions.

In order to make determinations happen faster with better accuracy, the team developed Real-Time Deep Memory InspectionTM (RTDMI), a patent-pending technology that allows malware to go straight to memory and extract the payload within the 100-nanosecond window it is exposed. The 2019 SonicWall Cyber Threat Report also mapped how the engine discovered nearly 75,000 ‘never-before-seen’ threats in 2018 alone — despite being released (at no additional cost to Capture ATP customers) in February 2018.

‘Never-Before-Seen’ Attacks Discovered by RTDMI in 2018

Image source: 2019 SonicWall Cyber Threat Report

Using proprietary machine learning capabilities, RTDMI has become more and more efficient at identifying and mitigating cyberattacks never seen by anyone in the cybersecurity industry. Since July 2018, the technology’s machine learning capabilities caught more undetectable cyberattacks in every month except one. In January 2019, this figure eclipsed 17,000 and continues to rise in 2019.

Year of the Processor Vulnerability

Much like how Heartbleed and other vulnerabilities in cryptographic libraries introduced researchers and attackers to a new battleground in 2014, so were the numerous announcements of vulnerabilities affecting processors in 2018.

Since these theoretical (currently) attacks operate in memory, RTDMI is well positioned to discover and stop these attacks from happening. By applying the information on how a theoretical attack would work to the machine learning engine, RTDMI was able to identify a Spectre attack within 30 days. Shortly thereafter, it was hardened for Meltdown. With each new processor vulnerability discovered (e.g., Foreshadow, PortSmash), it took RTDMI less and less time to harden against the attack.

Then, in March 2019, while much of the security world was at RSA Conference 2019 in San Francisco, the Spoiler vulnerability was announced. With the maturity found within RTDMI, it took the engine literally no time at all to identify if the vulnerability was being exploited.

Although we have yet to see these side-channel attacks in the wild, RTDMI is primed for the fight and even if there is a new vulnerability announced tomorrow with the ability to weaponize it, this layer of defense is ready to identify and block side-channel attacks against processor vulnerabilities.

Image source: 2019 SonicWall Cyber Threat Report

Scouting for New Technology

Now, if you are not a SonicWall customer yet and are evaluating solutions to stop unknown and ‘never-before-seen’ attacks (i.e., zero-day threats), ask your prospective vendors how they do against these types of attacks. Ask how they did on Day 1 of the WannaCry crisis. As for the volume of attacks their solutions are finding, ask for evidence the solution works in a real-world situation, not just as a proof of concept (POC) in a lab.

If you are a customer, Capture ATP, which includes RTDMI, is available as an add-on purchase within many of our offerings from the firewall, to email, to the wireless access point. You read that correctly: right on the access point.

We believe in the technology so much that we place it in everything to protect your networks and endpoints, such as laptops and IoT devices. This is why large enterprises, school districts, SMBs, retail giants, carrier networks and service providers, and government offices and agencies trust this technology to safeguard their networks, data and users every day.

How MSSPs & Artificial Intelligence Can Mitigate Zero-Day Threats

/0 Comments/in /by

So, here’s the problem: unknown zero-day threats are just that — unknown. You have no way (besides historical experience) to predict the next vulnerability avenue that will be exploited. You, therefore, don’t know what will need patching or what extra security layer needs injecting. This ultimately leads to a forecast-costing dilemma as you cannot predict the man hours involved.

The other quandary faced when tackling complex targeted zero days is the skills gap. Staffing a security operations center (SOC) with highly skilled cybersecurity professionals comes at a cost and only becomes profitable with economies of scale that a large customer base brings.

Coupled with the shortage of skilled cybersecurity professionals in the open market, how can you get your SOC off the ground? Could artificial intelligence (AI) level the playing field?

Machine Learning Reality Check

Machine learning and behavioral analytics continue to grow and become synonymous with zero-day threat protection. Is this all hype or is it the new reality? The truth is, it is both.

There is a lot of hype, but for good reason: AI works. Big data is needed to see the behaviors and therein the anomalies or outright nefarious activities that human oversight would mostly fail to catch. Delivered as a layered security approach, AI is the only way to truly protect against modern cyber warfare, but not all AI is deterministic and herein lies the hidden cost to your bottom line.

AI-based analysis tools that provide forensics are very powerful, but the horse has bolted by the time they are used. This approach is akin to intrusion detection systems (IDS) versus intrusion prevention systems (IPS). The former are great for retrospective audits, but what is the cleanup cost? This usage of behavioral analysis AI solely for detection is not MSSP-friendly. What you need is automated, real-time breach detection and prevention. Prevention is key.

So, how do you create an effective prevention technology? You need security layers that filter the malware noise, so each can be more efficient at its detection and prevention function than the last. That means signature-based solutions are still necessary. In fact, they are as important as ever as one of the first layers of defense in your arsenal (content filtering comes in at the top spot).

By SonicWall metrics, the ever-growing bombardment of attacks the average network faces stands at 1,200-plus per day (check out the mid-year update to the 2018 SonicWall Cyber Threat Report for more details).

When you do the math, it’s easy to see that with millions of active firewalls, it’s not practical to perform deep analysis on every payload. For the best results, you must efficiently fingerprint and filter everything that has gone before.

Aren’t All Sandboxes Basically the Same?

Only by understanding the behavior of the application and watching what it’s attempting to do, can you uncover malicious intent and criminal action. The best environment to do this is a sandbox, but no SOC manpower in the world could accomplish this with humans at scale. In order to be effective, you must turn to AI.

AI understands the big data coming from behavioral analysis. It can adapt the discovery approach to uncover threats that try to hide and, once determined as malicious, can fingerprint the payload via signature, turning a zero day into a known threat. It is the speed of propagation of this new, known signature to the protection appliances participating in the mesh protection network that drives the efficiencies to discover more threats.

Also, it’s the size of the mesh network catchment area that allows you the largest overall service area of attaches, which helps your AI quickly learn from the largest sample data set.

Luckily, SonicWall has you covered on all these fronts. With more than 1 million sensors deployed across 215 territories and countries, SonicWall has one of the largest global footprint of active firewalls. Plus, the cloud-based, multi-engine SonicWall Capture Advanced Threat Protection (ATP) sandbox service discovers and stops unknown, zero-day attacks, such as ransomware, at the gateway with automated remediation.

Our recent introduction of the patent-pending Real-Time Deep Memory Inspection (RTDMITM) technology, which inspects memory in real time, can detect and prevent chip vulnerability attaches such as Spectre, Meltdown and Foreshadow. It’s included with every Capture ATP activation.

At SonicWall, the mantra of automated, real-time breach detection and prevention is fundamental to our security portfolio. It is how our partners drive predictable operational expenditures in the most challenging security environments. Only via connected solutions, utilizing shared intelligence, can you protect against all cyber threat vectors.

A version of this story originally appeared on MSSP Alert and was republished with permission.

Foreshadow Vulnerability (L1TF) Introduces New Risks to Intel Processors

/0 Comments/in /by

A group of 10 threat researchers have disclosed a trio of new Spectre-based vulnerabilities that affect Intel chipsets. Named Foreshadow, the threats leverage a CPU design feature called speculative execution to defeat security controls used by Intel SGX (Software Guard eXtensions) processors.

“At its core, Foreshadow abuses a speculative execution bug in modern Intel processors, on top of which we develop a novel exploitation methodology to reliably leak plaintext enclave secrets from the CPU cache,” the research team published in its 18-page report Aug. 14.

The vulnerabilities are categorized as L1 Terminal Faults (L1TF). Intel published an overview, impact and mitigation guidance, and issued CVEs for each attack:

The research team found that Foreshadow abuses the same processor vulnerability as the Meltdown exploit, in which an attacker can leverage results of unauthorized memory accesses in transient out-of-order instructions before they are rolled back.

Conversely, Foreshadow uses a different attack model. Its goal is to “compromise state-of-the-art intra-address space enclave protection domains that are not covered by recently deployed kernel page table isolation defenses.”

“Once again, relentless researchers are demonstrating that cybercriminals can use the very architecture of processor chips to gain access to sensitive and often highly valued information,” said SonicWall President and CEO Bill Conner. “Like its predecessors Meltdown and Spectre, Foreshadow is attacking processor, memory and cache functions to extract sought after information. Once gained, side-channels can then be used to ‘pick locks’ within highly secured personal computers or even third-party clouds undetected.”

 

Does SonicWall protect customers from Foreshadow?

Yes. If a customer has the Capture Advanced Threat Protection (ATP) sandbox service activated, they are protected from current and future file-based Foreshadow exploits, as well as other chip-based exploits, via SonicWall’s patent-pended Real-Time Deep Memory Inspection (RTDMITM) technology.

“Fortunately, prior to Meltdown and Spectre being made public in January 2018, the SonicWall team was already developing Real-Time Deep Memory Inspection (RTDMITM) technology, which proactively protects customers against these very types of processor-based exploits, as well as PDF and Office exploits never before seen,” said Conner.

RTDMI is capable of detecting Foreshadow because RTDMI detection operates at the CPU instruction level and has full visibility into the code as the attack is taking place. This allows RTDMI to detect specific instruction permutations that lead to an attack.

“The guessed-at branch can cause data to be loaded into the cache, for example (or, conversely, it can push other data out of the cache),” explained Ars Technica technology editor Peter Bright. “These microarchitectural disturbances can be detected and measured — loading data from memory is quicker if it’s already in the cache.”

To be successful, cache timing must be “measured” by the attack or it can’t know what is or is not cached. This required measurement is detected by RTDMI and the attack is mitigated.

In addition, RTDMI can also detect this attack via its “Meltdown-style” exploit detection logic since user-level process will try to access privileged address space during attack execution.

Notice

SonicWall customers with the Capture Advanced Threat Protection (ATP) sandbox service activated are NOT vulnerable to file-based Foreshadow processor exploits.

How does Foreshadow impact my business, data or applications?

According to Intel’s official L1TF guidance, each variety of L1TF could potentially allow unauthorized disclosure of information residing in the SGX enclaves, areas of memory protected by the processor.

While no current real-world exploits are known, it’s imperative that organizations running virtual or cloud infrastructure, as well as those with sensitive workloads, apply microcode updates released by Intel (linked below) immediately. Meanwhile, SonicWall Capture Labs will continue to monitor the malware landscape in case these proofs of concept are weaponized.

“This class of attack is something that will not dissipate,” said Conner. “Instead, attackers will only seek to benefit from the plethora of malware strains available to them that they can formulate like malware cocktails to divert outdated technologies, security standards and tactics. SonicWall will continue to innovate and develop our threat detection and prevention arsenal so our customers can mitigate even the most historical of threats.”

What is speculative execution?

Speculative execution takes place when processors execute specific instructions ahead of time (as an optimization technique) before it is known that these instructions actually need to be executed. In conjunction with various branch-prediction algorithms, speculative execution enables significant improvement in processor performance.

What is L1 Terminal Fault?

Intel refers to a specific flaw that enables this class of speculative execution side-channel vulnerabilities as “L1 Terminal Fault” (L1TF). The flaw lies in permissions checking code terminating too soon when certain parts of the memory are (maliciously) marked in a certain manner.  For more information, please see Intel’s official definition and explanation of the L1TF vulnerability.

Are chips from other vendors at risk?

According to the research team, only Intel chips are affected by Foreshadow at this time.

What is Real-Time Deep Memory Inspection (RTDMI)?

RTDMI technology identifies and mitigates the most insidious cyber threats, including memory-based attacks. RTDMI proactively detects and blocks unknown mass-market malware — including malicious PDFs and attacks leveraging Microsoft Office documents — via deep memory inspection in real time.

“Our Capture Labs team has performed malware reverse-engineering and utilized machine learning for more than 20 years,” said Conner. “This research led to the development of RTDMI, which arms organizations to eliminate some of the biggest security challenges of all magnitudes, which now includes Foreshadow, as well as Meltdown and Spectre.”

RTDMI is a core multi-technology detection capability included in the SonicWall Capture ATP sandbox service. RTDMI identifies and blocks malware that may not exhibit any detectable malicious behavior or hides its weaponry via encryption.

To learn more, download the complimentary RTDMI solution brief.

How do I protect against Foreshadow vulnerability?

Please consult Intel’s official guidance and FAQ. To defend your organization against future processor-based attacks, including Foreshadow, Spectre and Meltdown, deploy a SonicWall next-generation firewall with an active Capture ATP sandbox license.

For small- and medium-sized businesses (SMB), also follow upcoming guidance provided via the new NIST Small Business Cybersecurity Act, which was signed into law on Aug. 14. The new policy “requires the Commerce Department’s National Institute of Standards and Technology to develop and disseminate resources for small businesses to help reduce their cybersecurity risks.”

NIST also offers a cybersecurity framework to help organizations of all sizes leverage best practices to better safeguard their networks, data and applications from cyberattacks.

Stop Memory-Based Attacks with Capture ATP

To mitigate file-based processor vulnerabilities like Meltdown, Spectre and Foreshadow, activate the Capture Advanced Threat Protection service with RTDMI. The multi-engine cloud sandbox proactively detects and blocks unknown mass-market malware and memory-based exploits like Foreshadow.

EXPLORE CAPTURE ATP