Category Archives: Microsoft Defender Advanced Threat Protection

MISA expands with new members and new product additions

Another RSA Conference (RSAC) and another big year for the Microsoft Intelligent Security Association (MISA). MISA was launched at RSAC 2018 with 26 members and a year later we had doubled in size to 53 members. Today, I am excited to share that the association has again doubled in size to 102 members.

New members expand the portfolio of MISA integrations

Our new members include a number of ecosystem partners, like RSA, ServiceNow, and Net Motion, which have developed critical integrations that benefit our shared customers and we look forward to deepening our relationship through MISA engagement.

New MISA member RSA is now using Azure Active Directory’s risky user data and other Microsoft security signals to enrich their risk score engine. Additionally, RSA also leverages the Graph Security API to feed their SIEM solution, RSA NetWitness with alerts from the entire suite of Microsoft Security solutions.

 “RSA is excited to showcase the RSA SecurID and RSA NetWitness integrations with Microsoft Security products. Our integrations with Microsoft Defender ATP, Microsoft Graph Security API, Azure AD, and Microsoft Azure Sentinel, help us to better secure access to our mutual customer’s applications, and detect threats and attacks. We’re excited to formalize the long-standing relationship through RSA Ready and MISA to better defend our customers against a world of increasing threats.”
—Anna Sarnek, Head of Strategic Business Development, Cloud and Identity for RSA

The ServiceNow Security Operations integration with Microsoft Graph Security API enables shared customers to automate incident management and response, leveraging the capabilities of the Now Platform’s single data model to dramatically improve their ability to prioritize and respond to threats generated by all Microsoft Security Solutions and custom alerts from Azure Sentinel.

“ServiceNow is pleased to join the Microsoft Intelligent Security Alliance to accelerate security incident response for our shared customers. The ServiceNow Security Operations integration with Azure Sentinel, via the graph security API, enables shared customers to automate incident management and response, leveraging the capabilities of the Now Platform’s single data model to dramatically improve their ability to prioritize and respond to threats.”
—Lou Fiorello, Head of Security Products for ServiceNow

Microsoft is pleased to welcome NetMotion, a connectivity and security solutions company for the world’s growing mobile workforce, into the security partner program. Using NetMotion’s class-leading VPN, customers not only gain uncompromised connectivity and feature parity, they benefit from a VPN that is compatible with Windows, MacOS, Android and iOS devices. For IT teams, NetMotion delivers visibility and control over the entire connection from endpoint to endpoint, over any network, through integration with Microsoft Endpoint Manager (Microsoft Intune).

“NetMotion is designed from the ground up to protect and enhance the user experience of any mobile device. By delivering plug-and-play integration with Microsoft Endpoint Manager, the mobile workforce can maximize productivity and impact without any disruption to their workflow from day one. For organizations already using or considering Microsoft, the addition of NetMotion’s VPN is an absolute no-brainer.”
—Christopher Kenessey, CEO of NetMotion Software

Expanded partner strategy for Microsoft Defender Advanced Threat Protection (ATP)

The Microsoft Defender ATP team worked with our ecosystem partners to take their rich and complete set of APIs a step further to extend the power of our combined platforms. This helps customers strengthen their network and endpoint security posture, add continuous security validation and attack simulation testing, orchestrate and automate incident correlation and remediation, and add threat intelligence and web content filtering capabilities. Read Extending Microsoft Defender ATP network of partners to learn more about their partner strategy expansion and their open framework philosophy.

New product teams join the association

In addition to growing our membership, MISA expanded to cover 12 of Microsoft’s security solutions, including our latest additions: Azure Security Center for IoT Security and Azure DDoS.

Azure Security Center for IoT Security announces five flagship integration partners

The simple onboarding flow for Azure Security Center for IoT enables you to protect your managed and unmanaged IoT devices, view all security alerts, reduce your attack surface with security posture recommendations, and run unified reports in a single pane of glass.

Through partnering with members like Attivo Networks, CyberMDX, CyberX, Firedome, and SecuriThings, Microsoft is able to leverage their vast knowledge pool to help customers defend against a world of increasing IoT threats in enterprise. These solutions protect managed and unmanaged IoT devices in manufacturing, energy, building management systems, healthcare, transportation, smart cities, smart homes, and more. Read more about IoT security and how these five integration partners are changing IoT security in this blog.

Azure DDoS Protection available to partners to combat DDoS attacks

The first DDoS attack occurred way back on July 22, 1999, when a network of 114 computers infected with a malicious script called Trin00 attacked a computer at the University of Minnesota, according to MIT Technology Review. Even after 20 years DDoS continues to be an ever-growing problem, with the number of DDoS attacks doubling in the last year alone and the types of attacks getting increasingly sophisticated with the explosion of IoT devices.

Azure DDoS Protection provides countermeasures against the most sophisticated DDoS threats. The service provides enhanced DDoS mitigation capabilities for your application and resources deployed in your virtual networks. Technology partners can now protect their customers’ resources natively with Azure DDoS Protection Standard to address the availability and reliability concerns due to DDoS attacks.

“Extending Azure DDoS Protection capabilities to Microsoft Intelligent Security Association will help our shared customers to succeed by leveraging the global scale of Azure Networking to protect their workloads against DDoS attacks”
—Anupam Vij, Principal Product Manager, Azure Networking

Learn more

To see MISA members in action, visit the Microsoft booth at RSA where we have a number of our security partners presenting and demoing throughout the week. To learn more about the Microsoft Intelligent Security Association, visit our webpage or the video playlist of member integrations. For more information on Microsoft security solutions, visit our website.

The post MISA expands with new members and new product additions appeared first on Microsoft Security.

Azure Sentinel uncovers the real threats hidden in billions of low fidelity signals

Cybercrime is as much a people problem as it is a technology problem. To respond effectively, the defender community must harness machine learning to compliment the strengths of people. This is the philosophy that undergirds Azure Sentinel. Azure Sentinel is a cloud-native SIEM that exploits machine learning techniques to empower security analysts, data scientists, and engineers to focus on the threats that matter. You may have heard of similar solutions from other vendors, but the Fusion technology that powers Azure Sentinel sets this SIEM apart for three reasons:

  1. Fusion finds threats that fly under the radar, by combining low fidelity, “yellow” anomalous activities into high fidelity “red” incidents.
  2. Fusion does this by using machine learning to combine disparate data—network, identity, SaaS, endpoint—from both Microsoft and Partner data sources.
  3. Fusion incorporates graph-based machine learning and a probabilistic kill chain to reduce alert fatigue by 90 percent.

Azure Sentinel

Intelligent security analytics for your entire enterprise.

Learn more

You can get a sense of how powerful Fusion is by looking at data from December 2019. During that month, billions of events flowed into Azure Sentinel from thousands of Azure Sentinel customers. Nearly 50 billion anomalous alerts were identified and graphed. After Fusion applied the probabilistic kill chain, the graph was reduced to 110 sub graphs. A second level of machine learning reduced it further to just 25 actionable incidents. This is how Azure Sentinel reduces alert fatigue by 90 percent.

Infographic showing alerts to high-fidelity incidents.

New Fusion scenarios—Microsoft Defender ATP + Palo Alto firewalls

There are currently 35 multi-stage attack scenarios generally available through Fusion machine learning technology in Azure Sentinel. Today, Microsoft has introduced several additional scenarios—in public preview—using Microsoft Defender Advanced Threat Protection (ATP) and Palo Alto logs. This way, you can leverage the power of Sentinel and Microsoft Threat Protection as complementary technologies for the best customer protection.

  • Detect otherwise missed attacks—By stitching together disparate datasets using Bayesian methods, Fusion helps to detect attacks that could have been missed.
  • Reduce mean time to remediate—Microsoft Threat Protection provides a best in class investigation experience when addressing alerts from Microsoft products. For non-Microsoft datasets, you can leverage hunting and investigation tools in Azure Sentinel.

Here are a few examples:

An endpoint connects to TOR network followed by suspicious activity on the Internal network—Microsoft Defender ATP detects that a user inside the network made a request to a TOR anonymization service. On its own this incident would be a low-level fidelity. It’s suspicious but doesn’t rise to the level of a high-level threat. Palo Alto firewalls registers anomalous activity from the same IP address, but it isn’t risky enough to block. Separately neither of these alerts get elevated, but together they indicate a multi-stage attack. Fusion makes the connection and promotes it to a high-fidelity incident.

Infographic of the Palo Alto firewall detecting threats.

A PowerShell program on an endpoint connects to a suspicious IP address, followed by suspicious activity on the Internal network—Microsoft Defender ATP generates an alert when a PowerShell program makes a suspicious network connection. If Palo Alto allows traffic from that IP address back into the network, Fusion ties the two incidents together to create a high-fidelity incident

An endpoint connects to a suspicious IP followed by anomalous activity on the Internal network—If Microsoft Defender ATP detects an outbound connection to an IP with a history of unauthorized access and Palo Alto firewalls allows an inbound request from that same IP address, it’s elevated by Fusion.

How Fusion works

  1. Construct graph

The process starts by collecting data from several data sources, such as Microsoft products, Microsoft security partner products, and other cloud providers. Each of those security products output anomalous activity, which together can number in the billions or trillions. Fusion gathers all the low and medium level alerts detected in a 30-day window and creates a graph. The graph is hyperconnected and consists of billions of vertices and edges. Each entity is represented by a vertex (or node). For example, a vertex could be a user, an IP address, a virtual machine (VM), or any other entity within the network. The edges (or links) represent all the activities. If a user accesses company resources with a mobile device, both the device and the user are represented as vertices connected by an edge.

Image of an AAD Detect graph.

Once the graph is built there are still billions of alerts—far too many for any security operations team to make sense of. However, within those connected alerts there may be a pattern that indicates something more serious. The human brain is just not equipped to quickly remove it. This is where machine learning can make a real difference.

  1. Apply probabilistic kill chain

Fusion applies a probabilistic kill chain which acts as a regularizer to the graph. The statistical analysis is based on how real people—Microsoft security experts, vendors, and customers—triage alerts. For example, defenders prioritize kill chains that are time bound. If a kill chain is executed within a day, it will take precedence over one that is enacted over a few days. An even higher priority kill chain is one in which all steps have been completed. This intelligence is encoded into the Fusion machine learning statistical model. Once the probabilistic kill chain is applied, Fusion outputs a smaller number of sub graphs, reducing the number of threats from billions to hundreds.

  1. Score the attack

To reduce the noise further, Fusion uses machine learning to apply a final round of scoring. If labeled data exists, Fusion uses random forests. Labeled data for attacks is generated from the extensive Azure red team that execute these scenarios. If labeled data doesn’t exist Fusion uses spectral clustering.

Some of the criteria used to elevate threats include the number of high impact activity in the graph and whether the subgraph connects to another subgraph.

The output of this machine learning process is tens of threats. These are extremely high priority alerts that require immediate action. Without Fusion, these alerts would likely remain hidden from view, since they can only be seen after two or more low level threats are stitched together to shine a light on stealth activities. AI-generated alerts can now be handed off to people who will determine how to respond.

The great promise of AI in cybersecurity is its ability to enable your cybersecurity people to stay one step ahead of the humans on the other side. AI-backed Fusion is just one example of the innovative potential of partnering technology and people to take on the threats of today and tomorrow.

Learn more

Read more about Azure Sentinel and dig into all the Azure Sentinel detection scenarios.

Also, bookmark the Security blog to keep up with our expert coverage on security matters. Follow us at @MSFTSecurity for the latest news and updates on cybersecurity.

The post Azure Sentinel uncovers the real threats hidden in billions of low fidelity signals appeared first on Microsoft Security.

Building on secure productivity

Among the most common and powerful attack vectors we have seen are those that exploit the daily tradeoff users make between security and productivity. Often, this can be as simple as a document hiding an exploit or a malicious link.

As an industry, we’re used to thinking of security and productivity in tension with each other. Security teams focus on blocking capabilities and reducing access to limit risk; users create workarounds or ignore policies to get their jobs done. Organizations may respond to increasing security threats by layering multiple security point solutions on top of each other, often increasing the complexity security teams manage while encouraging users to look for even more workarounds.

We don’t think this has to be the case.

Today, we‘re announcing two new Microsoft 365 capabilities that will help organizations stay both secure and productive at the same time. The power of these capabilities comes from the seamless integration between Windows 10, Office 365 ProPlus, and Microsoft Defender Advanced Threat Protection (ATP). We previously gave a “sneak peak” at Ignite and are excited to share publicly now.

Safe Documents is now available in public preview, rolling out over the next few days

With Safe Documents, we’re bringing the power of the Intelligent Security Graph down to the desktop to verify that documents are safe at the endpoint itself.

Although Protected View helps secure documents originating outside the organization, too often users would exit this sandbox without great consideration and leave their networks vulnerable. Bringing a minimal trust approach to the Office 365 ProPlus clients, Safe Documents automatically checks the document against known risks and threat profiles before allowing to open. Users are not asked to decide on their own whether a document can be trusted; they can simply focus on the work to be done. This seamless connection between the desktop and the cloud both simplifies the user workflow and helps to keep the network more secure.

Application Guard integration with Office 365 ProPlus is significantly expanding its private preview

With Application Guard, we created a micro-VM based on the same technology that powers the Azure cloud and brought it down to the desktop. We first introduced Application Guard in Edge, bringing hardware-level containerization to the browser.

Now integrated with Office 365 ProPlus, Application Guard provides an upgrade to Protected View that helps desktop users to stay safer and more productive with container-based isolation for Office applications. Application Guard’s enforcement—with a new instance of Windows 10 and separate copy of the kernel—completely blocks access to memory, local storage, installed applications, corporate network endpoints, or any other resources of interest to the attacker.

That means Office users will be able to open an untrusted Word, Excel, or PowerPoint file in a virtualized container. Users can stay productive—make edits, print, and save changes—all while protected with hardware-level security. If the untrusted file is malicious, the attack is contained while user data and identity remains untouched. When a user wants to trust a document to save on the network or start collaborating in real-time, Safe Documents will first check to help ensure the document is safe.

Moreover, both Safe Documents and Application Guard connect to the Microsoft Security Center, providing admins with advanced visibility and response capabilities including alerts, logs, confirmation the attack was contained, and the ability to see and act on similar threats across the enterprise.

Truly Microsoft 365 capabilities

With these new capabilities, we brought together some of the best of Windows 10, Office 365 ProPlus, and Microsoft Defender ATP to help organizations stay both secure and productive. This integration also means that organizations can deploy these features with the change of a setting and manage with existing tools. And with every malicious attack contained, the entire Intelligent Security Graph becomes stronger, benefiting everyone.

Both Safe Documents and Application Guard will be available to customers with Microsoft 365 E5 and E5 Security. We encourage customers to start testing Safe Documents in their environment as it comes available (initially available for tenants in the U.S., U.K., and European Union), and to learn more about Safe Documents and Application Guard.

The post Building on secure productivity appeared first on Microsoft Security.

Ghost in the shell: Investigating web shell attacks

Recently, an organization in the public sector discovered that one of their internet-facing servers was misconfigured and allowed attackers to upload a web shell, which let the adversaries gain a foothold for further compromise. The organization enlisted the services of Microsoft’s Detection and Response Team (DART) to conduct a full incident response and remediate the threat before it could cause further damage.

DART’s investigation showed that the attackers uploaded a web shell in multiple folders on the web server, leading to the subsequent compromise of service accounts and domain admin accounts. This allowed the attackers to perform reconnaissance using net.exe, scan for additional target systems using nbstat.exe, and eventually move laterally using PsExec.

The attackers installed additional web shells on other systems, as well as a DLL backdoor on an Outlook Web Access (OWA) server. To persist on the server, the backdoor implant registered itself as a service or as an Exchange transport agent, which allowed it to access and intercept all incoming and outgoing emails, exposing sensitive information. The backdoor also performed additional discovery activities as well as downloaded other malware payloads. In addition, the attackers sent special emails that the DLL backdoor interpreted as commands.

Figure 1. Sample web shell attack chain

The case is one of increasingly more common incidents of web shell attacks affecting multiple organizations in various sectors. A web shell is a piece of malicious code, often written in typical web development programming languages (e.g., ASP, PHP, JSP), that attackers implant on web servers to provide remote access and code execution to server functions. Web shells allow adversaries to execute commands and to steal data from a web server or use the server as launch pad for further attacks against the affected organization.

With the use of web shells in cyberattacks on the rise, Microsoft’s DART, the Microsoft Defender ATP Research Team, and the Microsoft Threat Intelligence Center (MSTIC) have been working together to investigate and closely monitor this threat.

Web shell attacks in the current threat landscape

Multiple threat actors, including ZINC, KRYPTON, and GALLIUM, have been observed utilizing web shells in their campaigns. To implant web shells, adversaries take advantage of security gaps in internet-facing web servers, typically vulnerabilities in web applications, for example CVE-2019-0604 or CVE-2019-16759.

In our investigations into these types of attacks, we have seen web shells within files that attempt to hide or blend in by using names commonly used for legitimate files in web servers, for example:

  • index.aspx
  • fonts.aspx
  • css.aspx
  • global.aspx
  • default.php
  • function.php
  • Fileuploader.php
  • help.js
  • write.jsp
  • 31.jsp

Among web shells used by threat actors, the China Chopper web shell is one of the most widely used. One example is written in JSP:

We have seen this malicious JSP code within a specially crafted file uploaded to web servers:

Figure 2. Specially crafted image file with malicious JSP code

Another China Chopper variant is written in PHP:

Meanwhile, the KRYPTON group uses a bespoke web shell written in C# within an ASP.NET page:

Figure 3. Web shell written in C# within an ASP.NET page

Once a web shell is successfully inserted into a web server, it can allow remote attackers to perform various tasks on the web server. Web shells can steal data, perpetrate watering hole attacks, and run other malicious commands for further compromise.

Web shell attacks have affected a wide range of industries. The organization in the public sector mentioned above represents one of the most common targeted sectors.

Aside from exploiting vulnerabilities in web applications or web servers, attackers take advantage of other weaknesses in internet-facing servers. These include the lack of the latest security updates, antivirus tools, network protection, proper security configuration, and informed security monitoring. Interestingly, we observed that attacks usually occur on weekends or during off-hours, when attacks are likely not immediately spotted and responded to.

Unfortunately, these gaps appear to be widespread, given that every month, Microsoft Defender Advanced Threat Protection (ATP) detects an average of 77,000 web shell and related artifacts on an average of 46,000 distinct machines.

Figure 3: Web shell encounters 

Detecting and mitigating web shell attacks

Because web shells are a multi-faceted threat, enterprises should build comprehensive defenses for multiple attack surfaces. Microsoft Threat Protection provides unified protection for identities, endpoints, email and data, apps, and infrastructure. Through signal-sharing across Microsoft services, customers can leverage Microsoft’s industry-leading optics and security technologies to combat web shells and other threats.

Gaining visibility into internet-facing servers is key to detecting and addressing the threat of web shells. The installation of web shells can be detected by monitoring web application directories for web script file writes. Applications such as Outlook Web Access (OWA) rarely change after they have been installed and script writes to these application directories should be treated as suspicious.

After installation, web shell activity can be detected by analyzing processes created by the Internet Information Services (IIS) process w3wp.exe. Sequences of processes that are associated with reconnaissance activity such as those identified in the alert screenshot (net.exe, ping.exe, systeminfo.exe, and hostname.exe) should be treated with suspicion. Web applications such as OWA run from well-defined Application Pools. Any cmd.exe process execution by w3wp.exe running from an application pool that doesn’t typically execute processes such as ‘MSExchangeOWAAppPool’ should be treated as unusual and regarded as potentially malicious.

Microsoft Defender ATP exposes these behaviors that indicate web shell installation and post-compromise activity by analyzing script file writes and process executions. When alerted of these activities, security operations teams can then use the rich capabilities in Microsoft Defender ATP to investigate and resolve web shell attacks.

Figure 4. Sample Microsoft Defender ATP alerts related to web shell attacks

Figure 5. Microsoft Defender ATP alert process tree

As in most security issues, prevention is critical. Organizations can harden systems against web shell attacks by taking these preventive steps:

  • Identify and remediate vulnerabilities or misconfigurations in web applications and web servers. Deploy latest security updates as soon as they become available.
  • Audit and review logs from web servers frequently. Be aware of all systems you expose directly to the internet.
  • Utilize the Windows Defender Firewall, intrusion prevention devices, and your network firewall to prevent command-and-control server communication among endpoints whenever possible. This limits lateral movement as well as other attack activities.
  • Check your perimeter firewall and proxy to restrict unnecessary access to services, including access to services through non-standard ports.
  • Enable cloud-delivered protection to get the latest defenses against new and emerging threats.
  • Educate end users about preventing malware infections. Encourage end users to practice good credential hygiene—limit the use of accounts with local or domain admin privileges.

 

 

Detection and Response Team (DART)

Microsoft Defender ATP Research Team

Microsoft Threat Intelligence Center (MSTIC)

 

The post Ghost in the shell: Investigating web shell attacks appeared first on Microsoft Security.

Threat hunting in Azure Advanced Threat Protection (ATP)

As members of Microsoft’s Detection and Response Team (DART), we’ve seen a significant increase in adversaries “living off the land” and using compromised account credentials for malicious purposes. From an investigation standpoint, tracking adversaries using this method is quite difficult as you need to sift through the data to determine whether the activities are being performed by the legitimate user or a bad actor. Credentials can be harvested in numerous ways, including phishing campaigns, Mimikatz, and key loggers.

Recently, DART was called into an engagement where the adversary had a foothold within the on-premises network, which had been gained through compromising cloud credentials. Once the adversary had the credentials, they began their reconnaissance on the network by searching for documents about VPN remote access and other access methods stored on a user’s SharePoint and OneDrive. After the adversary was able to access the network through the company’s VPN, they moved laterally throughout the environment using legitimate user credentials harvested during a phishing campaign.

Once our team was able to determine the initially compromised accounts, we were able to begin the process of tracking the adversary within the on-premises systems. Looking at the initial VPN logs, we identified the starting point for our investigation. Typically, in this kind of investigation, your team would need to dive deeper into individual machine event logs, looking for remote access activities and movements, as well as looking at any domain controller logs that could help highlight the credentials used by the attacker(s).

Luckily for us, this customer had deployed Azure Advanced Threat Protection (ATP) prior to the incident. By having Azure ATP operational prior to an incident, the software had already normalized authentication and identity transactions within the customer network. DART began querying the suspected compromised credentials within Azure ATP, which provided us with a broad swath of authentication-related activities on the network and helped us build an initial timeline of events and activities performed by the adversary, including:

  • Interactive logins (Kerberos and NTLM)
  • Credential validation
  • Resource access
  • SAMR queries
  • DNS queries
  • WMI Remote Code Execution (RCE)
  • Lateral Movement Paths

Azure Advanced Threat Protection

Detect and investigate advanced attacks on-premises and in the cloud.

Get started

This data enabled the team to perform more in-depth analysis on both user and machine level logs for the systems the adversary-controlled account touched. Azure ATP’s ability to identify and investigate suspicious user activities and advanced attack techniques throughout the cyber kill chain enabled our team to completely track the adversary’s movements in less than a day. Without Azure ATP, investigating this incident could have taken weeks—or even months—since the data sources don’t often exist to make this type of rapid response and investigation possible.

Once we were able to track the user throughout the environment, we were able to correlate that data with Microsoft Defender ATP to gain an understanding of the tools used by the adversary throughout their journey. Using the right tools for the job allowed DART to jump start the investigation; identify the compromised accounts, compromised systems, other systems at risk, and the tools being used by the adversaries; and provide the customer with the needed information to recover from the incident faster and get back to business.

Learn more and keep updated

Learn more about how DART helps customers respond to compromises and become cyber-resilient. Bookmark the Security blog to keep up with our expert coverage on security matters. Also, follow us at @MSFTSecurity for the latest news and updates on cybersecurity.

The post Threat hunting in Azure Advanced Threat Protection (ATP) appeared first on Microsoft Security.

CISO series: Lessons learned from the Microsoft SOC—Part 3b: A day in the life

The Lessons learned from the Microsoft SOC blog series is designed to share our approach and experience with security operations center (SOC) operations. We share strategies and learnings from our SOC, which protects Microsoft, and our Detection and Response Team (DART), who helps our customers address security incidents. For a visual depiction of our SOC philosophy, download our Minutes Matter poster.

For the next two installments in the series, we’ll take you on a virtual shadow session of a SOC analyst, so you can see how we use security technology. You’ll get to virtually experience a day in the life of these professionals and see how Microsoft security tools support the processes and metrics we discussed earlier. We’ll primarily focus on the experience of the Investigation team (Tier 2) as the Triage team (Tier 1) is a streamlined subset of this process. Threat hunting will be covered separately.

Image of security workers in an office.

General impressions

Newcomers to the facility often remark on how calm and quiet our SOC physical space is. It looks and sounds like a “normal” office with people going about their job in a calm professional manner. This is in sharp contrast to the dramatic moments in TV shows that use operations centers to build tension/drama in a noisy space.

Nature doesn’t have edges

We have learned that the real world is often “messy” and unpredictable, and the SOC tends to reflect that reality. What comes into the SOC doesn’t always fit into the nice neat boxes, but a lot of it follows predictable patterns that have been forged into standard processes, automation, and (in many cases) features of Microsoft tooling.

Routine front door incidents

The most common attack patterns we see are phishing and stolen credentials attacks (or minor variations on them):

  • Phishing email → Host infection → Identity pivot:

Infographic indicating: Phishing email, Host infection, and Identity pivot

  • Stolen credentials → Identity pivot → Host infection:

Infographic indicating: Stolen credentials, Identity pivot, and Host infection

While these aren’t the only ways attackers gain access to organizations, they’re the most prevalent methods mastered by most attackers. Just as martial artists start by mastering basic common blocks, punches, and kicks, SOC analysts and teams must build a strong foundation by learning to respond rapidly to these common attack methods.

As we mentioned earlier in the series, it’s been over two years since network-based detection has been the primary method for detecting an attack. We attribute this primarily to investments that improved our ability to rapidly remediate attacks early with host/email/identity detections. There are also fundamental challenges with network-based detections (they are noisy and have limited native context for filtering true vs. false positives).

Analyst investigation process

Once an analyst settles into the analyst pod on the watch floor for their shift, they start checking the queue of our case management system for incidents (not entirely unlike phone support or help desk analysts would).

While anything might show up in the queue, the process for investigating common front door incidents includes:

  1. Alert appears in the queue—After a threat detection tool detects a likely attack, an incident is automatically created in our case management system. The Mean Time to Acknowledge (MTTA) measurement of SOC responsiveness begins with this timestamp. See Part 1: Organization for more information on key SOC metrics.

Basic threat hunting helps keep a queue clean and tidy

Require a 90 percent true positive rate for alert sources (e.g., detection tools and types) before allowing them to generate incidents in the analyst queue. This quality requirement reduces the volume of false positive alerts, which can lead to frustration and wasted time. To implement, you’ll need to measure and refine the quality of alert sources and create a basic threat hunting process. A basic threat hunting process leverages experienced analysts to comb through alert sources that don’t meet this quality bar to identify interesting alerts that are worth investigating. This review (without requiring full investigation of each one) helps ensure that real incident detections are not lost in the high volume of noisy alerts. It can be a simple part time process, but it does require skilled analysts that can apply their experience to the task.

  1. Own and orient—The analyst on shift begins by taking ownership of the case and reading through the information available in the case management tool. The timestamp for this is the end of the MTTA responsiveness measurement and begins the Mean Time to Remediate (MTTR) measurement.

Experience matters

A SOC is dependent on the knowledge, skills, and expertise of the analysts on the team. The attack operators and malware authors you defend against are often adaptable and skilled humans, so no prescriptive textbook or playbook on response will stay current for very long. We work hard to take good care of our people—giving them time to decompress and learn, recruiting them from diverse backgrounds that can bring fresh perspectives, and creating a career path and shadowing programs that encourage them to learn and grow.

  1. Check out the host—Typically, the first priority is to identify affected endpoints so analysts can rapidly get deep insight. Our SOC relies on the Endpoint Detection and Response (EDR) functionality in Microsoft Defender Advanced Threat Protection (ATP) for this.

Why endpoint is important

Our analysts have a strong preference to start with the endpoint because:

  • Endpoints are involved in most attacks—Malware on an endpoint represents the sole delivery vehicle of most commodity attacks, and most attack operators still rely on malware on at least one endpoint to achieve their objective. We’ve also found the EDR capabilities detect advanced attackers that are “living off the land” (using tools deployed by the enterprise to navigate). The EDR functionality in Microsoft Defender ATP provides visibility into normal behavior that helps detect unusual command lines and process creation events.
  • Endpoint offers powerful insights—Malware and its behavior (whether automated or manual actions) on the endpoint often provides rich detailed insight into the attacker’s identity, skills, capabilities, and intentions, so it’s a key element that our analysts always check for.

Identifying the endpoints affected by this incident is easy for alerts raised by the Microsoft Defender ATP EDR, but may take a few pivots on an email or identity sourced alert, which makes integration between these tools crucial.

  1. Scope out and fill in the timeline—The analyst then builds a full picture and timeline of the related chain of events that led to the alert (which may be an adversary’s attack operation or false alarm positive) by following leads from the first host alert. The analyst travels along the timeline:
  • Backward in time—Track backward to identify the entry point in the environment.
  • Forward in time—Follow leads to any devices/assets an attacker may have accessed (or attempted to access).

Our analysts typically build this picture using the MITRE ATT&CK™ model (though some also adhere to the classic Lockheed Martin Cyber Kill Chain®).

True or false? Art or science?

The process of investigation is partly a science and partly an art. The analyst is ultimately building a storyline of what happened to determine whether this chain of events is the result of a malicious actor (often attempting to mask their actions/nature), a normal business/technical process, an innocent mistake, or something else.

This investigation is a repetitive process. Analysts identify potential leads based on the information in the original report, follow those leads, and evaluate if the results contribute to the investigation.

Analysts often contact users to identify whether they performed an anomalous action intentionally, accidentally, or was not done by them at all.

Running down the leads with automation

Much like analyzing physical evidence in a criminal investigation, cybersecurity investigations involve iteratively digging through potential evidence, which can be tedious work. Another parallel between cybersecurity and traditional forensic investigations is that popular TV and movie depictions are often much more exciting and faster than the real world.

One significant advantage of investigating cyberattacks is that the relevant data is already electronic, making it easier to automate investigation. For many incidents, our SOC takes advantage of security orchestration, automation, and remediation (SOAR) technology to automate investigation (and remediation) of routine incidents. Our SOC relies heavily on the AutoIR functionality in Microsoft Threat Protection tools like Microsoft Defender ATP and Office 365 ATP to reduce analyst workload. In our current configuration, some remediations are fully automatic and some are semi-automatic (where analysts review the automated investigations and propose remediation before approving execution of it).

Document, document, document

As the analyst builds this understanding, they must capture a complete record with their conclusions and reasoning/evidence for future use (case reviews, analyst self-education, re-opening cases that are later linked to active attacks, etc.).

As our analyst develops information on an incident, they capture the common, most relevant details quickly into the case such as:

  • Alert info: Alert links and Alert timeline
  • Machine info: Name and ID
  • User info
  • Event info
  • Detection source
  • Download source
  • File creation info
  • Process creation
  • Installation/Persistence method(s)
  • Network communication
  • Dropped files

Fusion and integration avoid wasting analyst time

Each minute an analyst wastes on manual effort is another minute the attacker has to spread, infect, and do damage during an attack operation. Repetitive manual activity also creates analyst toil, increases frustration, and can drive interest in finding a new job or career.

We learned that several technologies are key to reducing toil (in addition to automation):

  • Fusion—Adversary attack operations frequently trip multiple alerts in multiple tools, and these must be correlated and linked to avoid duplication of effort. Our SOC has found significant value from technologies that automatically find and fuse these alerts together into a single incident. Azure Security Center and Microsoft Threat Protection include these natively.
  • Integration—Few things are more frustrating and time consuming than having to switch consoles and tools to follow a lead (a.k.a., swivel chair analytics). Switching consoles interrupts their thought process and often requires manual tasks to copy/paste information between tools to continue their work. Our analysts are extremely appreciative of the work our engineering teams have done to bring threat intelligence natively into Microsoft’s threat detection tools and link together the consoles for Microsoft Defender ATP, Office 365 ATP, and Azure ATP. They’re also looking forward to (and starting to test) the Microsoft Threat Protection Console and Azure Sentinel updates that will continue to reduce the swivel chair analytics.

Stay tuned for the next segment in the series, where we’ll conclude our investigation, remediate the incident, and take part in some continuous improvement activities.

Learn more

In the meantime, bookmark the Security blog to keep up with our expert coverage on security matters and follow us at @MSFTSecurity for the latest news and updates on cybersecurity.

To learn more about SOCs, read previous posts in the Lessons learned from the Microsoft SOC series, including:

Watch the CISO Spotlight Series: Passwordless: What’s It Worth.

Also, see our full CISO series and download our Minutes Matter poster for a visual depiction of our SOC philosophy.

The post CISO series: Lessons learned from the Microsoft SOC—Part 3b: A day in the life appeared first on Microsoft Security.

Mobile threat defense and intelligence are a core part of cyber defense

The modern workplace is a mobile workplace. Today’s organizations rely on mobility to increase productivity and improve the customer experience. But the proliferation of smartphones and other mobile devices has also expanded the attack surface of roughly 5 billion mobile devices in the world, many used to handle sensitive corporate data. To safeguard company assets, organizations need to augment their global cyber defense strategy with mobile threat intelligence.

When handled and analyzed properly, actionable data holds the key to enabling solid, 360-degree cybersecurity strategies and responses. However, many corporations lack effective tools to collect, analyze, and act on the massive volume of security events that arise daily across their mobile fleet. An international bank recently faced this challenge. By deploying Pradeo Security alongside Microsoft Endpoint Manager and Microsoft Defender Advanced Threat Protection (ATP), the bank was able to harness its mobile data and better protect the company.

Pradeo Security strengthens Microsoft Endpoint Manager Conditional Access policies

In 2017, the Chief Information Security Office (CISO) of an international bank recognized that the company needed to address the risk of data exposure on mobile. Cybercriminals exploit smart phones at the application, network, and OS levels, and infiltrate them through mobile applications 78 percent of the time.1 The General Data Protection Regulation (GDPR) was also scheduled to go into effect the following year. The company needed to better secure its mobile data to safeguard the company and comply with the new privacy regulations.

The company deployed Microsoft Endpoint Manager to gain visibility into the mobile devices accessing corporate resources. Microsoft Endpoint Manager is the recently announced convergence of Microsoft Intune and Configuration Manager functionality and data, plus new intelligent actions, offering seamless, unified endpoint management. Then, to ensure the protection of these corporate resources, the company deployed Pradeo Security Mobile Threat Defense, which is integrated with Microsoft.

Pradeo Security and Microsoft Endpoint Manager work together to apply conditional access policies to each mobile session. Conditional access policies allow the security team to automate access based on the circumstances. For example, if a user tries to gain access using a device that is not managed by Microsoft Endpoint Manager, the user may be forced to enroll the device. Pradeo Security enhances Microsoft Endpoint Manager’s capabilities by providing a clear security status of any mobile devices accessing corporate data, which Microsoft can evaluate for risk. If a smartphone is identified as non-compliant based on the data that Pradeo provides, conditional access policies can be applied.

For example, if the risk is high, the bank could set policies that block access. The highly granular and customizable security policies offered by Pradeo Security gave the CISO more confidence that the mobile fleet was better protected against threats specifically targeting his industry.

Get more details about Pradeo Security for Microsoft Endpoint Manager in this datasheet.

Detect and respond to advanced cyberthreats with Pradeo Security and Microsoft Defender ATP

The bank also connected Pradeo Security to Microsoft Defender ATP in order to automatically feed it with always current mobile security inputs. Microsoft Defender ATP helps enterprises prevent, detect, investigate, and respond to advanced cyberthreats. Pradeo Security enriches Microsoft Defender ATP with mobile security intelligence. Immediately, the bank was able to see information on the latest threats targeting their mobile fleet. Only a few weeks later, there was enough data in the Microsoft platform to draw trends and get a clear understanding of the company’s mobile threat environment.

Pradeo relies on a network of millions of devices (iOS and Android) across the globe to collect security events related to the most current mobile threats. Pradeo leverages machine learning mechanisms to distill and classify billions of raw and anonymous security facts into actionable mobile threat intelligence.

Today, this bank’s mobile ecosystem entirely relies on Pradeo and Microsoft, as its security team finds it to be the most cost-effective combination when it comes to mobile device management, protection, and intelligence.

About Pradeo

Pradeo is a global leader of mobile security and a member of the Microsoft Intelligent Security Association (MISA). It offers services to protect the data handled on mobile devices and applications, and tools to collect, process, and get value out of mobile security events.

Pradeo’s cutting-edge technology has been recognized as one of the most advanced mobile security technologies by Gartner, IDC, and Frost & Sullivan. It provides a reliable detection of mobile threats to prevent breaches and reinforce compliance with data privacy regulations.

For more details, contact Pradeo.

Note: Users must be entitled separately to Pradeo and Microsoft licenses as appropriate.

Learn more

To learn more about MISA, visit the MISA webpage. Also, bookmark the Security blog to keep up with our expert coverage on security matters and follow us at @MSFTSecurity for the latest news and updates on cybersecurity.

Microsoft Endpoint Manager

Transformative management and security that meets you where you are and helps you move to the cloud.

Get started

12019 Mobile Security Report, Pradeo Lab

The post Mobile threat defense and intelligence are a core part of cyber defense appeared first on Microsoft Security.

Data science for cybersecurity: A probabilistic time series model for detecting RDP inbound brute force attacks

Computers with Windows Remote Desktop Protocol (RDP) exposed to the internet are an attractive target for adversaries because they present a simple and effective way to gain access to a network. Brute forcing RDP, a secure network communications protocol that provides remote access over port 3389, does not require a high level of expertise or the use of exploits; attackers can utilize many off-the-shelf tools to scan the internet for potential victims and leverage similar such tools for conducting the brute force attack.

Attackers target RDP servers that use weak passwords and are without multi-factor authentication, virtual private networks (VPNs), and other security protections. Through RDP brute force, threat actor groups can gain access to target machines and conduct many follow-on activities like ransomware and coin mining operations.

In a brute force attack, adversaries attempt to sign in to an account by effectively using one or more trial-and-error methods. Many failed sign-ins occurring over very short time frequencies, typically minutes or even seconds, are usually associated with these attacks. A brute force attack might also involve adversaries attempting to access one or more accounts using valid usernames that were obtained from credential theft or using common usernames like “administrator”. The same holds for password combinations. In detecting RDP brute force attacks, we focus on the source IP address and username, as password data is not available.

In the Windows operating system, whenever an attempted sign-in fails for a local machine, Event Tracing for Windows (ETW) registers Event ID 4625 with the associated username. Meanwhile, source IP addresses connected to RDP can be accessed; this information is very useful in assessing if a machine is under brute force attack. Using this information in combination with Event ID 4624 for non-server Windows machines can shed light on which sign-in sessions were successfully created and can further help in detecting if a local machine has been compromised.

In this blog we’ll present a study and a detection logic that uses these signals. This data science-driven approach to detecting RDP brute force attacks has proven valuable in detecting human adversary activity through Microsoft Threat Experts, the managed threat hunting service in Microsoft Defender Advanced Threat Protection. This work is an example of how the close collaboration between data scientists and threat hunters results in protection for customers against real-world threats.

Insights into brute force attacks

Observing a sudden, relatively large count of Event ID 4625 associated with RDP network connections might be rare, but it does not necessarily imply that a machine is under attack. For example, a script that performs the following actions would look suspicious looking at a time series of counts of failed sign-in but is most likely not malicious:

  • uses an expired password
  • retries sign-in attempts every N-minutes with different usernames
  • over a public IP address within a range owned by the enterprise

In contrast, behavior that includes the following is indicative of an attack:

  • extreme counts of failed sign-ins from many unknown usernames
  • never previously successfully authenticated
  • from multiple RDP connections
  • from new source IP addresses

Understanding the context of failed sign-ins and inbound connections is key to discriminating between true positive (TP) and false positive (FP) brute force attacks, especially if the goal is to automatically raise only high-precision alerts to the appropriate recipients, as we do in Microsoft Defender ATP.

We analyzed several months’ worth of data to mine insights into the types of RDP brute force attacks occurring across Microsoft Defender ATP customers. Out of about 45,000 machines that had both RDP public IP connections and at least 1 network failed sign-in, we discovered that, on average, several hundred machines per day had high probability of undergoing one or more RDP brute force attack attempts. Of the subpopulation of machines with detected brute force attacks, the attacks lasted 2-3 days on average, with about 90% of cases lasting for 1 week or less, and less than 5% lasting for 2 weeks or more.

Figure 1: Empirical distribution in number of days per machine where we observed 1 or more brute force attacks

As discussed in numerous other studies [1], large counts of failed sign-ins are often associated with brute force attacks. Looking at the count of daily failed sign-ins, 90% of cases exceeded 10 attempts, with a median larger than 60. In addition, these unusual daily counts had high positive correlation with extreme counts in shorter time windows (see Figure 2). In fact, the number of extreme failed sign-ins per day typically occurred under 2 hours, with about 40% failing in under 30 minutes.

Figure 2: Count of daily and maximum hourly network failed sign-ins for a local machine under brute force attack

While a detection logic based on thresholding the count of failed sign-ins during daily or finer grain time window can detect many brute force attacks, this will likely produce too many false positives. Worse, relying on just this will yield false negatives, missing successful enterprise compromises: our analysis revealed several instances where brute force attacks generated less than 5-10 failed attempts at a daily granularity but often persisted for many days, thereby avoiding extreme counts at any point in time. For such a brute force attack, thresholding the cumulative number of failed sign-ins across time could be more useful, as depicted in Figure 3.

Figure 3: Daily and cumulative failed network sign-in

Looking at counts of network failed sign-ins provides a useful but incomplete picture of RDP brute force attacks. This can be further augmented with additional information on the failed sign-in, such as the failure reason, time of day, and day of week, as well as the username itself. An especially strong signal is the source IP of the inbound RDP connection. Knowing if the external IP has a high reputation of abuse, as can be looked up on sites like https://www.abuseipdb.com/, can directly confirm if an IP is a part of an active brute force.

Unfortunately, not all IP addresses have a history of abuse; in addition, it can be expensive to retrieve information about many external IP addresses on demand. Maintaining a list of suspicious IPs is an option, but relying on this can result in false negatives as, inevitably, new IPs continually occur, particularly with the adoption of cloud computing and ease of spinning up virtual machines. A generic signal that can augment failed sign-in and user information is counting distinct RDP connections from external IP addresses. Again, extreme values occurring at a given time or cumulated over time can be an indicator of attack.

Figure 4 shows histograms (i.e., counts put into discrete bins) of daily counts of RDP public connections per machine that occurred for an example enterprise with known brute force attacks. It’s evident that normal machines have a lower probability of larger counts compared to machines attacked.

Figure 4: Histograms of daily count of RDP inbound across machines for an example enterprise

Given that some enterprises have machines under brute force attack daily, the priority may be to focus on machines that have been compromised, defined by a first successful sign-in following failed attempts from suspicious source IP addresses or unusual usernames. In Windows logs, Event ID 4624 can be leveraged to measure successful sign-in events for local machine in combination with failed sign-ins (Event ID 4625).

Out of the hundreds of machines with RDP brute force attacks detected in our analysis, we found that about .08% were compromised. Furthermore, across all enterprises analyzed over several months, on average about 1 machine was detected with high probability of being compromised resulting from an RDP brute force attack every 3-4 days. Figure 5 shows a bubble chart of the average abuse score of external IPs associated with RDP brute force attacks that successfully compromised machines. The size of the bubbles is determined by the count of distinct machines across the enterprises analyzed having a network connection from each IP. While there is diversity in the origin of the source IPs, Netherlands, Russia, and the United Kingdom have a larger concentration of inbound RDP connections from high-abuse IP.

Figure 5: Bubble chart of IP abuse score versus counts of machine with inbound RDP

A key takeaway from our analysis is that successful brute force attempts are not uncommon; therefore, it’s critical to monitor at least the suspicious connections and unusual failed sign-ins that result in authenticated sign-in events. In the following sections we describe a methodology to do this. This methodology was leveraged by Microsoft Threat Experts to augment threat hunting and resulted in new targeted attack notifications.

Combining many relevant signals

As discussed earlier (with the example of scripts connecting via RDP using outdated passwords yielding failed sign-ins), simply relying on thresholding failed attempts per machine for detecting brute force attacks can be noisy and may result in many false positives. A better strategy is to utilize many contextually relevant signals, such as:

  • the timing, type, and count of failed sign-in
  • username history
  • type and frequency of network connections
  • first-time username from a new source machine with a successful sign-in

This can be even further extended to include indicators of attack associated with brute force, such as port scanning.

Combining multiple signals along the attack chain has been proposed and shown promising results [2]. We considered the following signals in detecting RDP inbound brute force attacks per machine:

  • hour of day and day of week of failed sign-in and RDP connections
  • timing of successful sign-in following failed attempts
  • Event ID 4625 login type (filtered to network and remote interactive)
  • Event ID 4625 failure reason (filtered to %%2308, %%2312, %%2313)
  • cumulative count of distinct username that failed to sign in without success
  • count (and cumulative count) of failed sign-ins
  • count (and cumulative count) of RDP inbound external IP
  • count of other machines having RDP inbound connections from one or more of the same IP

Unsupervised probabilistic time series anomaly detection

For many cybersecurity problems, including detecting brute force attacks, previously labeled data is not usually available. Thus, training a supervised learning model is not feasible. This is where unsupervised learning is helpful, enabling one to discover and quantify unknown behaviors when examples are too sparse. Given that several of the signals we consider for modeling RDP brute force attacks are inherently dependent on values observed over time (for example, daily counts of failed sign-ins and counts of inbound connections), time series models are particularly beneficial. Specifically, time series anomaly detection naturally provides a logical framework to quantify uncertainty in modeling temporal changes in data and produce probabilities that then can be ranked and thresholded to control a desirable false positive rate.

Time series anomaly detection captures the temporal dynamics of signals and accurately quantifies the probability of observing values at any point in time under normal operating conditions. More formally, if we introduce the notation Y(t) to denote the signals taking on values at time t, then we build a model to compute reliable estimates of the probability of Y(t) exceeding observed values given all known and relevant information, represented by P[y(t)], sometimes called an anomaly score. Given a false positive tolerance rate r (e.g., .1% or 1 out of 10,000 per time), for each time t, values y*(t) satisfying P[y*(t)] < r would be detected as anomalous. Assuming the right signals reflecting the relevant behaviors of the type of attacks are chosen, then the idea is simple: the lowest anomaly scores occurring per time will be likely associated with the highest likelihood of real threats.

For example, looking back at Figure 2, the time series of daily count of failed sign-ins occurring on the brute force attack day 8/4/2019 had extreme values that would be associated with an empirical probability of about .03% out of all machine and days with at least 1 failed network sign-in for the enterprise.

As discussed earlier, applying anomaly detection to 1 or a few signals to detect real attacks can yield too many false positives. To mitigate this, we combined anomaly scores across eight signals we selected to model RDP brute force attack patterns. The details of our solution are included in the Appendix, but in summary, our methodology involves:

  • updating statistical discrete time series models sequentially for each signal, capturing time of day, day of week, and both point and cumulative effects
  • combining anomaly scores using an approach that yields accurate probability estimates, and
  • ranking the top N anomalies per day to control a desired number of false positives

Our approach to time series anomaly detection is computationally efficient, automatically learns how to update probabilities and adapt to changes in data.

As we describe in the next section, this approach has yielded successful attack detection at high precision.

Protecting customers from real-word RDP brute force attacks through Microsoft Threat Experts

The proposed time series anomaly detection model was deployed and utilized by Microsoft Threat Experts to detect RDP brute force attacks during threat hunting activities. A list that ranks machines across enterprises with the lowest anomaly scores (indicating the likelihood of observing a value at least as large under expected conditions in all signals considered) is updated and reviewed every day. See Table 1 for an example.

Table 1: Sample ranking of detected RDP inbound brute force attacks

For each machine with detection of a probable brute force attack, each instance is assigned TP, FP, or unknown. Each TP is then assigned priority based on the severity of the attack. For high-priority TP, a targeted attack notification is sent to the associated organization with details about the active brute force attack and recommendations for mitigating the threat; otherwise the machine is closely monitored until more information is available.

We also added an extra capability to our anomaly detection: automatically sending targeted attack notifications about RDP brute force attacks, in many cases before the attack succeeds or before the actor is able to conduct further malicious activities. Looking at the most recent sample of about two weeks of graded detections, the average precision per day (i.e., true positive rate) is approximately 93.7% at a conservative false positive rate of 1%.

In conclusion, based on our careful selection of signals found to be highly associated with RDP brute force attacks, we demonstrated that proper application of time series anomaly detection can be very accurate in identifying real threats. We have filed a patent application for this probabilistic time series model for detecting RDP inbound brute force attacks. In addition, we are working on integrating this capability into Microsoft Defender ATP’s endpoint and detection response capabilities so that the detection logic can raise alerts on RDP brute force attacks in real-time.

Monitoring suspicious activity in failed sign-ins and network connections should be taken seriously—a real-time anomaly detection capable of self-updating with the changing dynamics in a network can indeed provide a sustainable solution. While Microsoft Defender ATP already has many anomaly detection capabilities integrated into its EDR capabilities, which enrich advanced threat protection across the broader Microsoft Threat Protection, we will continue to enhance these detections to cover more security scenarios. Using data science, we will continue to combine robust statistical and machine learning approaches with threat expertise and intelligence to deliver industry-leading protection to our customers through Microsoft Threat Protection.

 

 

Cole Sodja, Justin Carroll, Joshua Neil
Microsoft Defender ATP Research Team

 

 

Appendix 1: Models formulation

We utilize hierarchical zero-adjusted negative binomial dynamic models to capture the characteristics of the highly discrete count time series. Specifically, as shown in Figure 2, it’s expected that most of the time there won’t be failed sign-ins for valid credentials on a local machine; hence, there are excess zeros that would not be explained by standard probability distributions such as the negative binomial. In addition, the variance of non-zero counts is often much larger than the mean, where for example, valid scripts connecting via RDP can generate counts in the 20s or more over several minutes because of an outdated password. Moreover, given a combination of multiple users or scripts connecting to shared machines at the same time, this can generate more extreme counts at higher quantiles resulting in heavier tails, as seen in Figure 6.

Figure 6: Daily count of network failed sign-in for a machine with no brute force attack

Parametric discrete location/scale distributions do not generate well-calibrated p-values for rare time series, as seen in Figure 6, and thus if used to detect anomalies can result in too many FPs when looking across many machines at high time frequencies. To overcome this challenge dealing with the sparse time series of counts of failed sign-in and RDP inbound public connections we specify a mixture model, where, based on our analysis, a zero-inflated two-component negative binomial distribution was adequate.

Our formulation is based on thresholding values that determine when to transition to a distribution with larger location and/or scale as given in Equation 1. Hierarchical priors are given from empirical estimates of the sample moments across machines using about 1 month of data.

Equation 1: Zero-adjusted negative binomial threshold model

Negative binomial distribution (NB):

To our knowledge, this formulation does not yield a conjugate prior, and so directly computing probabilities from the posterior predicted density is not feasible. Instead, anomaly scores are generated based on drawing samples from all distributions and then computing the empirical right-tail p-value.

Updating parameters is done based on applying exponential smoothing. To avoid outliers skewing estimates, such as machines under brute force or other attacks, trimming is applied to sample from the distribution at a specified false positive rate, which was set to .1% for our study. Algorithm 1 outlines the logic.

The smoothing parameters were learned based on maximum likelihood estimation and then fixed during each new sequential update. To induce further uncertainty, bootstrapping across machines is done to produce a histogram of smoothing weights, and samples are drawn in accordance to their frequency. We found that weights concentrated away from 0 vary between .06% and 8% for over 90% of machines, thus leading to slow changes in the parameters. An extension using adaptive forgetting factors will be considered in future work to automatically learn how to correct smoothing in real time.

Algorithm 2: Updating model parameters real-time

Appendix 2: Fisher Combination

For a given device, for each signal that exists a score is computed defined as a p-value, where lower values are associated with higher likelihood of being an anomaly. Then the p-values are combined to yield a joint score across all signals based on using the Fisher p-value combination method as follows:

The use of Fisher’s test applied to anomaly scores produces a scalable solution that yields interpretable probabilities that thus can be controlled to achieve a desired false positive rate. This has even been applied in a cybersecurity context. [3]

 

 

[1] Najafabadi et al, Machine Learning for Detecting Brute Force Attacks at the Network Level, 2014 IEEE 14th International Conference on Bioinformatics and Bioengineering
[2] Sexton et al, Attack chain detection, Statistical Analysis and Data Mining, 2015
[3] Heard, Combining Weak Statistical Evidence in Cyber Security, Intelligent Data Analysis XIV, 2015

 

 

 


Talk to us

Questions, concerns, or insights on this story? Join discussions at the Microsoft Defender ATP community.

Read all Microsoft security intelligence blog posts.

Follow us on Twitter @MsftSecIntel.

The post Data science for cybersecurity: A probabilistic time series model for detecting RDP inbound brute force attacks appeared first on Microsoft Security.