Skip to content

Posts tagged ‘aws cybersecurity’

10 Ways AI And Machine Learning Are Improving Endpoint Security

  • Gartner predicts $137.4B will be spent on Information Security and Risk Management in 2019, increasing to $175.5B in 2023, reaching a CAGR of 9.1%. Cloud Security, Data Security, and Infrastructure Protection are the fastest-growing areas of security spending through 2023.
  •  69% of enterprise executives believe artificial intelligence (AI) will be necessary to respond to cyberattacks with the majority of telecom companies (80%) saying they are counting on AI to help identify threats and thwart attacks according to Capgemini.
  •  Spending on AI-based cybersecurity systems and services reached $7.1B in 2018 and is predicted to reach $30.9B in 2025, attaining a CAGR of 23.4% in the forecast period according to Zion Market Research.

Traditional approaches to securing endpoints based on the hardware characteristics of a given device aren’t stopping breach attempts today. Bad actors are using AI and machine learning to launch sophisticated attacks to shorten the time it takes to compromise an endpoint and successfully breach systems. They’re down to just 7 minutes after comprising an endpoint and gaining access to internal systems ready to exfiltrate data according to Ponemon. The era of trusted and untrusted domains at the operating system level, and “trust, but verify” approaches are over. Security software and services spending is soaring as a result, as the market forecasts above show.

AI & Machine Learning Are Redefining Endpoint Security

AI and machine learning are proving to be effective technologies for battling increasingly automated, well-orchestrated cyberattacks and breach attempts. Attackers are combining AI, machine learning, bots, and new social engineering techniques to thwart endpoint security controls and gain access to enterprise systems with an intensity never seen before. It’s becoming so prevalent that Gartner predicts that more than 85% of successful attacks against modern enterprise user endpoints will exploit configuration and user errors by 2025. Cloud platforms are enabling AI and machine learning-based endpoint security control applications to be more adaptive to the proliferating types of endpoints and corresponding threats. The following are the top ten ways AI and machine learning are improving endpoint security:

  • Using machine learning to derive risk scores based on previous behavioral patterns, geolocation, time of login, and many other variables is proving to be effective at securing and controlling access to endpoints. Combining supervised and unsupervised machine learning to fine-tune risk scores in milliseconds is reducing fraud, thwarting breach attempts that attempt to use privileged access credentials, and securing every identity on an organizations’ network. Supervised machine learning models rely on historical data to find patterns not discernable with rules or predictive analytics. Unsupervised machine learning excels at finding anomalies, interrelationships, and valid links between emerging factors and variables. Combining both unsupervised and supervised machine learning is proving to be very effective in spotting anomalous behavior and reducing or restricting access.
  • Mobile devices represent a unique challenge to achieving endpoint security control, one that machine learning combined with Zero Trust is proving to be integral at solving.  Cybercriminals prefer to steal a mobile device, its passwords, and privileged access credentials than hack into an organization. That’s because passwords are the quickest onramp they have to the valuable data they want to exfiltrate and sell. Abandoning passwords for new techniques including MobileIron’s zero sign-on approach shows potential for thwarting cybercriminals from getting access while hardening endpoint security control. Securing mobile devices using a zero-trust platform built on a foundation of unified endpoint management (UEM) capabilities enables enterprises to scale zero sign-on for managed and unmanaged services for the first time. Below is a graphic illustrating how they’re adopting machine learning to improve mobile endpoint security control:
  • Capitalizing on the core strengths of machine learning to improve IT asset management is making direct contributions to greater security.  IT Management and security initiatives continue to become more integrated across organizations, creating new challenges to managing endpoint security across each device. Absolute Software is taking an innovative approach to solve the challenge of improving IT asset management, so endpoint protection is strengthened at the same time. Recently I had a chance to speak with Nicko van Someren, Ph.D. and Chief Technology Officer at Absolute Software, where he shared with me how machine learning algorithms are improving security by providing greater insights into asset management. “Keeping machines up to date is an IT management job, but it’s a security outcome. Knowing what devices should be on my network is an IT management problem, but it has a security outcome. And knowing what’s going on and what processes are running and what’s consuming network bandwidth is an IT management problem, but it’s a security outcome. I don’t see these as distinct activities so much as seeing them as multiple facets of the same problem space. Nicko added that Absolute’s endpoint security controls begin at the BIOS level of over 500M devices that have their endpoint code embedded in them. The Absolute Platform is comprised of three products: Persistence, Intelligence, and Resilience—each building on the capabilities of the other. Absolute Intelligence standardizes the data around asset analytics and security advocacy analytics to allow Security managers to ask any question they want. (“What’s slowing down my device? What’s working and what isn’t? What has been compromised? What’s consuming too much memory? How does this deviate from normal performance?”). An example of Absolute’s Intelligence providing insights into asset management and security is shown below:
  • Machine learning has progressed to become the primary detection method for identifying and stopping malware attacks. Machine learning algorithms initially contributed to improving endpoint security by supporting the back-end of malware protection workflows. Today more vendors are designing endpoint security systems with machine learning as the primary detection method. Machine learning trained algorithms can detect file-based malware and learn which files are harmful or not based on the file’s metadata and content. Symantec’s Content & Malware Analysis illustrates how machine learning is being used to detect and block malware. Their approach combines advanced machine learning and static code file analysis to block, detect, and analyze threats and stop breach attempts before they can spread.
  • Supervised machine learning algorithms are being used for determining when given applications are unsafe to use, assigning them to containers, so they’re isolated from production systems. Taking into account an applications’ threat score or reputation, machine learning algorithms are defining if dynamic application containment needs to run for a given application. Machine learning-based dynamic application containment algorithms and rules block or log unsafe actions of an application based on containment and security rules. Machine learning algorithms are also being used for defining predictive analytics that define the extent of a given applications’ threat.
  •  Integrating AI, machine learning, and SIEM (Security Information and Event Management) in a single unified platform are enabling organizations to predict, detect, and respond to anomalous behaviors and events. AI and machine learning-based algorithms and predictive analytics are becoming a core part of SIEM platforms today as they provide automated, continuous analysis and correlation of all activity observed within a given IT environment. Capturing, aggregating, and analyzing endpoint data in real-time using AI techniques and machine learning algorithms is providing entirely new insights into asset management and endpoint security. One of the most interesting companies to watch in this area is LogRhythm. They’ve developed an innovative approach to integrating AI, machine learning, and SIEM in their LogRhythm NextGen SIEM Platform, which delivers automated, continuous analysis and correlation of all activity observed within an IT environment. The following is an example of how LogRhythm combines AI, machine learning, and SIEM to bring new insights into securing endpoints across a network.
  • Machine learning is automating the more manually-based, routine incident analysis, and escalation tasks that are overwhelming security analysts today. Capitalizing on supervised machine learnings’ innate ability to fine-tune algorythms in milliseconds based on the analysis of incidence data, endpoint security providers are prioritizing this area in product developnent. Demand from potential customers remains strong, as nearly everyone is facing a cybersecurity skills shortage while facing an onslaught of breach attempts.  “The cybersecurity skills shortage has been growing for some time, and so have the number and complexity of attacks; using machine learning to augment the few available skilled people can help ease this. What’s exciting about the state of the industry right now is that recent advances in Machine Learning methods are poised to make their way into deployable products,” Absolute’s CTO Nicko van Someren added.
  • Performing real-time scans of all processes with an unknown or suspicious reputation is another way how machine learning is improving endpoint security. Commonly referred to as Hunt and Respond, supervised and unsupervised machine learning algorithms are being used today to seek out and resolve potential threats in milliseconds instead of days. Supervised machine learning algorithms are being used to discover patterns in known or stable processes where anomalous behavior or activity will create an alert and pause the process in real-time. Unsupervised machine learning algorithms are used for analyzing large-scale, unstructured data sets to categorize suspicious events, visualize threat trends across the enterprise, and take immediate action at a single endpoint or across the entire organization.
  • Machine learning is accelerating the consolidation of endpoint security technologies, a market dynamic that is motivating organizations to trim back from the ten clients they have on average per endpoint today. Absolute Software’s 2019 Endpoint Security Trends Report found that a typical device has ten or more endpoint security agents installed, each often conflicting with the other. The study also found that enterprises are using a diverse array of endpoint agents, including encryption, AV/AM, and Endpoint Detection and Response (EDR). The wide array of endpoint solutions make it nearly impossible to standardize a specific test to ensure security and safety without sacrificing speed. By helping to accelerate the consolidation of security endpoints, machine learning is helping organizations to see the more complex and layered the endpoint protection, the greater the risk of a breach.
  • Keeping every endpoint in compliance with regulatory and internal standards is another area machine learning is contributing to improving endpoint security. In regulated industries, including financial services, insurance, and healthcare, machine learning is being deployed to discover, classify, and protect sensitive data. This is especially the case with HIPAA (Health Insurance Portability and Accountability Act) compliance in healthcare. Amazon Macie is representative of the latest generation of machine learning-based cloud security services. Amazon Macie recognizes sensitive data such as personally identifiable information (PII) or intellectual property and provides organizations with dashboards, alerts, and contextual insights that give visibility into how data is being accessed or moved. The fully managed service continuously monitors data access activity for anomalies and generates detailed alerts when it detects the risk of unauthorized access or inadvertent data leaks. An example of one of Amazon Macie’s dashboard is shown below:

The Truth About Privileged Access Security On AWS And Other Public Clouds

 

Bottom Line: Amazon’s Identity and Access Management (IAM) centralizes identity roles, policies and Config Rules yet doesn’t go far enough to provide a Zero Trust-based approach to Privileged Access Management (PAM) that enterprises need today.

AWS provides a baseline level of support for Identity and Access Management at no charge as part of their AWS instances, as do other public cloud providers. Designed to provide customers with the essentials to support IAM, the free version often doesn’t go far enough to support PAM at the enterprise level. To AWS’s credit, they continue to invest in IAM features while fine-tuning how Config Rules in their IAM can create alerts using AWS Lambda. AWS’s native IAM can also integrate at the API level to HR systems and corporate directories, and suspend users who violate access privileges.

In short, native IAM capabilities offered by AWS, Microsoft Azure, Google Cloud, and more provides enough functionality to help an organization get up and running to control access in their respective homogeneous cloud environments. Often they lack the scale to fully address the more challenging, complex areas of IAM and PAM in hybrid or multi-cloud environments.

The Truth about Privileged Access Security on Cloud Providers Like AWS

The essence of the Shared Responsibility Model is assigning responsibility for the security of the cloud itself including the infrastructure, hardware, software, and facilities to AWS and assign the securing of operating systems, platforms, and data to customers. The AWS version of the Shared Responsibility Model, shown below, illustrates how Amazon has defined securing the data itself, management of the platform, applications and how they’re accessed, and various configurations as the customers’ responsibility:

AWS provides basic IAM support that protects its customers against privileged credential abuse in a homogenous AWS-only environment. Forrester estimates that 80% of data breaches involve compromised privileged credentials, and a recent survey by Centrify found that 74% of all breaches involved privileged access abuse.

The following are the four truths about privileged access security on AWS (and, generally, other public cloud providers):

  1. Customers of AWS and other public cloud providers should not fall for the myth that cloud service providers can completely protect their customized and highly individualized cloud instances. As the Shared Responsibility Model above illustrates, AWS secures the core areas of their cloud platform, including infrastructure and hosting services. AWS customers are responsible for securing operating systems, platforms, and data and most importantly, privileged access credentials. Organizations need to consider the Shared Responsibility Model the starting point on creating an enterprise-wide security strategy with a Zero Trust Security framework being the long-term goal. AWS’s IAM is an interim solution to the long-term challenge of achieving Zero Trust Privilege across an enterprise ecosystem that is going to become more hybrid or multi-cloud as time goes on.
  2. Despite what many AWS integrators say, adopting a new cloud platform doesn’t require a new Privileged Access Security model. Many organizations who have adopted AWS and other cloud platforms are using the same Privileged Access Security Model they have in place for their existing on-premises systems. The truth is the same Privileged Access Security Model can be used for on-premises and IaaS implementations. Even AWS itself has stated that conventional security and compliance concepts still apply in the cloud. For an overview of the most valuable best practices for securing AWS instances, please see my previous post, 6 Best Practices For Increasing Security In AWS In A Zero Trust World.
  3. Hybrid cloud architectures that include AWS instances don’t need an entirely new identity infrastructure and can rely on advanced technologies, including Multi-Directory Brokering. Creating duplicate identities increases cost, risk, and overhead and the burden of requiring additional licenses. Existing directories (such as Active Directory) can be extended through various deployment options, each with their strengths and weaknesses. Centrify, for example, offers Multi-Directory Brokering to use whatever preferred directory already exists in an organization to authenticate users in hybrid and multi-cloud environments. And while AWS provides key pairs for access to Amazon Elastic Compute Cloud (Amazon EC2) instances, their security best practices recommend a holistic approach should be used across on-premises and multi-cloud environments, including Active Directory or LDAP in the security architecture.
  4. It’s possible to scale existing Privileged Access Management systems in use for on-premises systems today to hybrid cloud platforms that include AWS, Google Cloud, Microsoft Azure, and other platforms. There’s a tendency on the part of system integrators specializing in cloud security to oversell cloud service providers’ native IAM and PAM capabilities, saying that a hybrid cloud strategy requires separate systems. Look for system integrators and experienced security solutions providers who can use a common security model already in place to move workloads to new AWS instances.

Conclusion

The truth is that Identity and Access Management solutions built into public cloud offerings such as AWS, Microsoft Azure, and Google Cloud are stop-gap solutions to a long-term security challenge many organizations are facing today. Instead of relying only on a public cloud provider’s IAM and security solutions, every organization’s cloud security goals need to include a holistic approach to identity and access management and not create silos for each cloud environment they are using. While AWS continues to invest in their IAM solution, organizations need to prioritize protecting their privileged access credentials – the “keys to the kingdom” – that if ever compromised would allow hackers to walk in the front door of the most valuable systems an organization has. The four truths defined in this article are essential for building a Zero Trust roadmap for any organization that will scale with them as they grow. By taking a “never trust, always verify, enforce least privilege” strategy when it comes to their hybrid- and multi-cloud strategies, organizations can alleviate costly breaches that harm the long-term operations of any business.

%d bloggers like this: