Tag Archives: Threat Intelligence

Introducing Cloudflare One Intel

Post Syndicated from Malavika Balachandran Tadeusz original https://blog.cloudflare.com/cloudflare-one-intel/

Introducing Cloudflare One Intel

Introducing Cloudflare One Intel

Earlier this week, we announced Cloudflare One, a single platform for networking and security management. Cloudflare One extends the speed, reliability, and security we’ve brought to Internet properties and applications over the last decade to make the Internet the new enterprise WAN.

Underpinning Cloudflare One is Cloudflare’s global network – today, our network spans more than 200 cities worldwide and is within milliseconds of nearly everyone connected to the Internet. Our network handles, on average, 18 million HTTP requests and 6 million DNS requests per second. With 1 billion unique IP addresses connecting to the Cloudflare network each day, we have one of the broadest views on Internet activity worldwide.

We see a large diversity of Internet traffic across our entire product suite. Every day, we block 72 billion cyberthreats. This visibility provides us with a unique position to understand and mitigate Internet threats, and enables us to see new threats and malware before anyone else.

At the beginning of this month, as part of our 10th Birthday Week, we launched Cloudflare Radar, which shares high-level trends with the general public based on our network’s aggregate data. The same data that powers that view of the Internet also gives us the ability to create new insights to keep your team safer.

Today, we are excited to announce the next phase of network and threat intelligence at Cloudflare: the launch of Cloudflare One Intel. Cloudflare One Intel streamlines network and security operations by converting the data we can gather on our network into actionable insights.

The challenge with the traditional security operations

Most enterprises use a large array of point solutions to ensure that the corporate network remains fast, available and secure. Security teams typically aggregate logs from these point solutions into their SIEM and create custom alerts for incident detection.

Once an incident has been detected, security teams will quickly respond with remediating actions to prevent data loss, such as removing a compromised device’s access controls or adding a malicious hostname or URL to a block list.

Along with incident remediation, security teams will conduct an investigation of the incident to uncover more details about the attacker. Pivoting across historical DNS records, SSL certificate fingerprints, malware samples, and other indicators of compromise, security researchers will try to uncover more details about an attacker. Linked indicators then get fed back onto block lists in point solutions to prevent subsequent attacks.

However, there are several challenges with traditional incident detection and response. Security operations teams are often overwhelmed by the plethora of logs and alerts. With threat intelligence, SIEMs, and control planes all in different platforms, incident detection, remediation and forensics can be slow, arduous, and expensive.

Improving Incident Response with Cloudflare One

We want to make network and security operations as streamlined as possible. Cloudflare One Intel helps network and security teams detect and respond to incidents more efficiently. That means bringing together insights from your network activity, global Internet intelligence, and automated remediation in a single platform.

As part of the mission to help security teams detect and block emerging security threats more efficiently we are releasing two features within Cloudflare Gateway: DNS tunneling detection and domain insights.

What is DNS Tunneling?

DNS tunneling is the misuse of the Domain Name System (DNS) protocol to encode another protocol’s data into a series of DNS queries and response messages. DNS tunneling is often used to circumvent a corporate firewall. For example, DNS tunneling might be used to visit a website that is blocked on the corporate firewall, distribute malware from a command & control server, or exfiltrate sensitive data.

DNS tunneling isn’t only used for malicious activities. One of the most common uses of DNS tunneling is by antivirus software, which will often use DNS tunneling to look up file signatures.

Blocking DNS tunneling using Cloudflare Gateway

Starting today, customers using Cloudflare Gateway can block hostnames associated with DNS tunneling using the “DNS Tunneling” filter in Gateway’s DNS filtering policies. This feature is available to all Gateway users at no additional cost.

You can begin using the filter by navigating to the Policies section of the Gateway product and selecting the “Security Threats” tab. Once you check the “DNS Tunneling” box, Gateway will automatically block any requests made by your organization’s users to domains on this list. Should you want to manually override any specific domains, you can use the “Domain Override” feature to remove the block policy on a specific domain.

Introducing Cloudflare One Intel

We previously included known malicious DNS tunnels in our “Anonymizer” category within Gateway’s security threat categories. We are now pulling that into its own category so that customers can have more granular visibility into threats on their network. Further, we are expanding the filter beyond known malicious DNS tunnels to include newly emerging threats, so that customers can block these threats as soon as we see them on our network.

How we use machine learning to detect DNS tunneling

Using machine learning, Cloudflare detects anomalous DNS request patterns and flags these requests as suspected DNS tunneling. Our model analyzes requests and detects anomalous behavior at a frequency of every five minutes.

Once a set of requests is flagged, we add the associated hostname to our “DNS Tunneling” category. We do not add hostnames of commonly allowed DNS tunnels to this list, such as those used by antivirus software.

Our model not only blocks hostnames associated with DNS tunneling seen on your network, but across the entire Cloudflare network. Processing over 500 billion DNS queries each day, we have unique insight into global DNS traffic patterns.

Adding transparency to security

Cloudflare’s unique insight into global Internet traffic is what powers the intelligence behind Cloudflare One. DNS tunneling detection is one example of how we use aggregated data from our network to improve Internet security for everyone. But, until now, that has been opaque to users.

Security teams investigating the threats that impact their organization need more transparency. Cloudflare One Intel consolidates the information we have about the potentially harmful sites and properties that can target your organization.

Starting today, with a single click, administrators reviewing logs in Cloudflare Gateway can get a comprehensive breakdown of any site being allowed or blocked.

In this expanded view, you can now click the “View Domain Insights” button, which will take you to the Cloudflare Radar Domain Insights page for the requested hostname. This feature is available to all Gateway users at no additional cost.

Introducing Cloudflare One Intel
Introducing Cloudflare One Intel

What’s Next

These new features are just the beginning of Cloudflare One Intel. Over the coming weeks and months, we’ll be rolling out more features across the Cloudflare One platform that will make our Internet intelligence more accessible and actionable. Stay tuned for premium features available in Cloudflare Radar for Cloudflare Gateway customers.

Get started now

Cloudflare Radar is available to everyone for free – you can check it out here and start exploring our Internet intelligence.

To protect your team from threats on the Internet that utilize DNS tunnelling, sign up for a Cloudflare Gateway account and use the Security filter setting to block DNS tunnelling attempts. DNS-based security and content filtering is available for free across every Gateway plan.

How Security Operation Centers can use Amazon GuardDuty to detect malicious behavior

Post Syndicated from Darren House original https://aws.amazon.com/blogs/security/how-security-operation-centers-can-use-amazon-guardduty-to-detect-malicious-behavior/

The Security Operations Center (SOC) has a tough job. As customers modernize and shift to cloud architectures, the ability to monitor, detect, and respond to risks poses different challenges.

In this post we address how Amazon GuardDuty can address some common concerns of the SOC regarding the number of security tools and the overhead to integrate and manage them. We describe the GuardDuty service, how the SOC can use GuardDuty threat lists, filtering, and suppression rules to tune detections and reduce noise, and the intentional model used to define and categorize GuardDuty finding types to quickly give detailed information about detections.

Today, the typical SOC has between 10 and 60 tools for managing security. Some larger SOCs can have more than 100 tools, which are mostly point solutions that don’t integrate with each other.

The security market is flush with niche security tools you can deploy to monitor, detect, and respond to events. Each tool has technical and operational overhead in the form of designing system uptime, sensor deployment, data aggregation, tool integration, deployment plans, server and software maintenance, and licensing.

Tuning your detection systems is an example of a process with both technical and operational overhead. To improve your signal-to-noise ratio (S/N), the security systems you deploy have to be tuned to your environment and to emerging risks that are relevant to your environment. Improving the S/N matters for SOC teams because it reduces time and effort spent on activities that don’t bring value to an organization. Spending time tuning detection systems reduces the exhaustion factors that impact your SOC analysts. Tuning is highly technical, yet it’s also operational because it’s a process that continues to evolve, which means you need to manage the operations and maintenance lifecycle of the infrastructure and tools that you use in tuning your detections.

Amazon GuardDuty

GuardDuty is a core part of the modern FedRAMP-authorized cloud SOC, because it provides SOC analysts with a broad range of cloud-specific detective capabilities without requiring the overhead associated with a large number of security tools.

GuardDuty is a continuous security monitoring service that analyzes and processes data from Amazon Virtual Private Cloud (VPC) Flow Logs, AWS CloudTrail event logs that record Amazon Web Services (AWS) API calls, and DNS logs to provide analysis and detection using threat intelligence feeds, signatures, anomaly detection, and machine learning in the AWS Cloud. GuardDuty also helps you to protect your data stored in S3. GuardDuty continuously monitors and profiles S3 data access events (usually referred to as data plane operations) and S3 configurations (control plane APIs) to detect suspicious activities. Detections include unusual geo-location, disabling of preventative controls such as S3 block public access, or API call patterns consistent with an attempt to discover misconfigured bucket permissions. For a full list of GuardDuty S3 threat detections, see GuardDuty S3 finding types. GuardDuty integrates threat intelligence feeds from CrowdStrike, Proofpoint, and AWS Security to detect network and API activity from known malicious IP addresses and domains. It uses machine learning to identify unknown and potentially unauthorized and malicious activity within your AWS environment.

The GuardDuty team continually monitors and manages the tuning of detections for threats related to modern cloud deployments, but your SOC can use trusted IP and threat lists and suppression rules to implement your own custom tuning to fit your unique environment.

GuardDuty is integrated with AWS Organizations, and customers can use AWS Organizations to associate member accounts with a GuardDuty management account. AWS Organizations helps automate the process of enabling and disabling GuardDuty simultaneously across a group of AWS accounts that are in your control. Note that, as of this writing, you can have one management account and up to 5,000 member accounts.

Operational overhead is near zero. There are no agents or sensors to deploy or manage. There are no servers to build, deploy, or manage. There’s nothing to patch or upgrade. There aren’t any highly available architectures to build. You don’t have to buy a subscription to a threat intelligence provider, manage the influx of threat data and most importantly, you don’t have to invest in normalizing all of the datasets to facilitate correlation. Your SOC can enable GuardDuty with a single click or API call. It is a multi-account service where you can create a management account, typically in the security account, that can read all findings information from the member accounts for easy centralization of detections. When deployed in a Management/Member design, GuardDuty provides a flexible model for centralizing your enterprise threat detection capability. The management account can control certain member settings, like update intervals for Amazon CloudWatch Events, use of threat and trusted lists, creation of suppression rules, opening tickets, and automating remediations.

Filters and suppression rules

When GuardDuty detects potential malicious activity, it uses a standardized finding format to communicate the details about the specific finding. The details in a finding can be queried in filters, displayed as saved rules, or used for suppression to improve visibility and reduce analyst fatigue.

Suppress findings from vulnerability scanners

A common example of tuning your GuardDuty deployment is to use suppression rules to automatically archive new Recon:EC2/Portscan findings from vulnerability assessment tools in your accounts. This is a best practice designed to reduce S/N and analyst fatigue.

The first criteria in the suppression rule should use the Finding type attribute with a value of Recon:EC2/Portscan. The second filter criteria should match the instance or instances that host these vulnerability assessment tools. You can use the Instance image ID attribute, the Network connection remote IPv4 address, or the Tag value attribute depending on what criteria is identifiable with the instances that host these tools. In the example shown in Figure 1, we used the remote IPv4 address.

Figure 1: GuardDuty filter for vulnerability scanners

Figure 1: GuardDuty filter for vulnerability scanners

Filter on activity that was not blocked by security groups or NACL

If you want visibility into the GuardDuty detections that weren’t blocked by preventative measures such as a network ACL (NACL) or security group, you can filter by the attribute Network connection blocked = False, as shown in Figure 2. This can provide visibility into potential changes in your filtering strategy to reduce your risk.

Figure 2: GuardDuty filter for activity not blocked by security groups on NACLs

Figure 2: GuardDuty filter for activity not blocked by security groups on NACLs

Filter on specific malicious IP addresses

Some customers want to track specific malicious IP addresses to see whether they are generating findings. If you want to see whether a single source IP address is responsible for CloudTrail-based findings, you can filter by the API caller IPv4 address attribute.

Figure 3: GuardDuty filter for specific malicious IP address

Figure 3: GuardDuty filter for specific malicious IP address

Filter on specific threat provider

Maybe you want to know how many findings are generated from a threat intelligence provider or your own threat lists. You can filter by the attribute Threat list name to see if the potential attacker is on a threat list from CrowdStrike, Proofpoint, AWS, or your threat lists that you uploaded to GuardDuty.

Figure 4: GuardDuty filter for specific threat intelligence list provider

Figure 4: GuardDuty filter for specific threat intelligence list provider

Finding types and formats

Now that you know a little more about GuardDuty and tuning findings by using filters and suppression rules, let’s dive into the finding types that are generated by a GuardDuty detection. The first thing to know is that all GuardDuty findings use the following model:


ThreatPurpose:ResourceTypeAffected/ThreatFamilyName.ThreatFamilyVariant!Artifact

This model is intended to communicate core information to security teams on the nature of the potential risk, the AWS resource types that are potentially impacted, and the threat family name, variants, and artifacts of the detected activity in your account. The Threat Purpose field describes the primary purpose of a threat or a potential attempt on your environment.

Let’s take the Backdoor:EC2/C&CActivity.B!DNS finding as an example.


ThreatPurpose:ResourceTypeAffected/ThreatFamilyName.ThreatFamilyVariant!Artifact
Backdoor     :EC2                 /C&CActivity.    .B                  !DNS

The Backdoor threat purpose indicates an attempt to bypass normal security controls on a specific Amazon Elastic Compute Cloud (EC2) instance. In this example, the EC2 instance in your AWS environment is querying a domain name (DNS) associated with a known command and control (C&CActivity) server. This is a high-severity finding, because there are enough indicators that malware is on your host and acting with malicious intent.

GuardDuty, at the time of this writing, supports the following finding types:

  • Backdoor finding types
  • Behavior finding types
  • CryptoCurrency finding types
  • PenTest finding types
  • Persistence finding types
  • Policy finding types
  • PrivilegeEscalation finding types
  • Recon finding types
  • ResourceConsumption finding types
  • Stealth finding types
  • Trojan finding types
  • Unauthorized finding types

OK, now you know about the model for GuardDuty findings, but how does GuardDuty work?

When you enable GuardDuty, the service evaluates events in both a stateless and stateful manner, which allows us to use machine learning and anomaly detection in addition to signatures and threat intelligence. Some stateless examples include the Backdoor:EC2/C&CActivity.B!DNS or the CryptoCurrency:EC2/BitcoinTool.B finding types, where GuardDuty only needs to see a single DNS query, VPC Flow Log entry, or CloudTrail record to detect potentially malicious activity.

Stateful detections are driven by anomaly detection and machine learning models that identify behaviors that deviate from a baseline. The machine learning detections typically require more time to train the models and potentially use multiple events for triggering the detection.

The typical triggers for behavioral detections vary based on the log source and the detection in question. The behavioral detections learn about typical network or user activity to set a baseline, after which the anomaly detections change from a learning mode to an active mode. In active mode, you only see findings generated from these detections if the service observes behavior that suggests a deviation. Some examples of machine learning–based detections include the Backdoor:EC2/DenialOfService.Dns, UnauthorizedAccess:IAMUser/ConsoleLogin, and Behavior:EC2/NetworkPortUnusual finding types.

Why does this matter?

We know the SOC has the tough job of keeping organizations secure with limited resources, and with a high degree of technical and operational overhead due to a large portfolio of tools. This can impact the ability to detect and respond to security events. For example, CrowdStrike tracks the concept of breakout time—the window of time from when an outside party first gains unauthorized access to an endpoint machine, to when they begin moving laterally across your network. They identified average breakout times are between 19 minutes and 10 hours. If the SOC is overburdened with undifferentiated technical and operational overhead, it can struggle to improve monitoring, detection, and response. To act quickly, we have to decrease detection time and the overhead burden on the SOC caused by the numerous tools used. The best way to decrease detection time is with threat intelligence and machine learning. Threat intelligence can provide context to alerts and gives a broader perspective of cyber risk. Machine learning uses baselines to detect what normal looks like, enabling detection of anomalies in user or resource behavior, and heuristic threats that change over time. The best way to reduce SOC overhead is to share the load so that AWS services manage the undifferentiated heavy lifting, while the SOC focuses on more specific tasks that add value to the organization.

GuardDuty is a cost-optimized service that is in scope for the FedRAMP and DoD compliance programs in the US commercial and GovCloud Regions. It leverages threat intelligence and machine learning to provide detection capabilities without you having to manage, maintain, or patch any infrastructure or manage yet another security tool. With a 30-day trial period, there is no risk to evaluate the service and discover how it can support your cyber risk strategy.

If you want to receive automated updates about GuardDuty, you can subscribe to an SNS notification that will email you whenever new features and detections are released.

If you have feedback about this post, submit comments in the Comments section below. If you have questions about this post, start a new thread on the Amazon GuardDuty forum or contact AWS Support.

Want more AWS Security how-to content, news, and feature announcements? Follow us on Twitter.

Author

Darren House

Darren brings over 20 years’ experience building secure technology architectures and technical strategies to support customer outcomes. He has held several roles including CTO, Director of Technology Solutions, Technologist, Principal Solutions Architect, and Senior Network Engineer for USMC. Today, he is focused on enabling AWS customers to adopt security services and automations that increase visibility and reduce risk.

Author

Trish Cagliostro

Trish is a leader in the security industry where she has spent 10 years advising public and private sector customers like DISA, DHS, and US Senate and commercial entities like Bank of America and United Airlines. Trish is a subject matter expert on a variety of topics, including integrating threat intelligence and has testified before the House Homeland Security Committee about information sharing.