Showing posts with label cybersecurity. Show all posts
Showing posts with label cybersecurity. Show all posts

Tuesday, October 15, 2024

Getting Your SOC SOARing Despite AI

It’s a fact: enterprise security operations centers (SOCs) that are most satisfied with their investments in Security Information and Event Management (SIEM) and Security Orchestration, Automation, and Response (SOAR) operate and maintain less than a dozen playbooks. This is something I’ve uncovered in recent years whilst building SIEM+SOAR and autonomous SOC solutions – and it perhaps runs counterintuitive to many security leaders’ visions for SOAR use and value.

SOAR technology is one of those much-touted security silver bullets that have tarnished over time and been subsumed into broader categories of threat detection and incident response (TDIR) solutions, yet it continues to remain a distinct must-have modern SOC capability.


Why do satisfied SOCs run so few playbooks? After all, a core premise of SOAR was that you can automate any (perhaps all) security responses, and your SOC analysts would spend less time on ambient security noise – giving them more time to focus on higher-priority incidents. Surely “success” would include automating as many responses as possible?

Beyond the fact that “security is hard,” the reality is that threat detection and response is as dynamic as the organization you’re trying to protect. New systems, new policies, new business owners, new tools, and a sea of changing security products and API connector updates mean that playbooks must be dynamic and vigilantly maintained, or they become stale, broken, and ineffective.

Every SOC team has at least one playbook covering their phishing response. It’s one of the most common and frequently encountered threats within the enterprise, yet “phishing” covers an amazingly broad range of threats and possible responses, so a playbook-based response to the threat is programmatically very complex and brittle to environmental changes.

From a SOC perspective of automating and orchestrating a response, you would either build a lengthy single if/then/else-stye playbook or craft individual playbooks for each permutation of the threat. Smart SOC operators quickly learn that the former is more maintainable and scalable than the latter. A consequence of this is that you need analysts with more experience to maintain and operate the playbook. Any analyst can knock-up a playbook for a simple or infrequently encountered threat vector, but it takes business knowledge and vigilance to maintain each playbook’s efficacy beyond the short term.

Surely AI and a sprinkle of LLM magic will save the day though, right?

I can’t count the number of security vendors and startups that have materialized over the last couple of years with AI and LLM SOAR capabilities, features, or solutions – all with pitches that suggest playbooks are dead, dying, replaced, accelerated, automatically maintained, dynamically created, managed, open sourced, etc., so the human SOC analyst does less work going forward. I remain hopeful that 10% of any of that eventually becomes true.

For the immediate future, SOC teams should continue to be wary of any AI stories that emphasize making it easier to create playbooks (or their product-specific equivalent of a playbook). More is NOT better. It’s too easy (and all too common) to fall down the rathole of creating a new playbook for an existing threat because it’s too hard to find and maintain an earlier iteration of that threat’s playbook. Instead, focus on a subset of the most common and time-consuming threats that your SOC already faces daily, and nail them using the smallest number of playbooks you can get away with.

With the Rolling Stones’ “(I Can’t Get No) Satisfaction” playing in the background, perhaps you’ll get some modicum of SIEM+SOAR satisfaction by keeping your playbook playlist under a dozen.

-- Gunter Ollmann

First Published: IOActive Blog - October 15, 2024

Tuesday, August 20, 2024

Get Strategic About Cyber Risk Management

With global cybercrime damage costs exceeding $11 trillion last year and moving toward an estimated $20 trillion by 2026, robust cybersecurity risk management has never been more imperative.


The interconnected nature of modern technology means that, by default, even small vulnerabilities can lead to catastrophic losses. And it’s not just about finances. Unmitigated risk raises the specter of eroded customer confidence and tainted brand reputation. In this comprehensive guide, we’ll give enterprise defenders a holistic, methodical, checklist-style approach to cybersecurity risk management. We’ll focus on practical applications, best practices, and ready-to-implement strategies designed to mitigate risks and safeguard digital assets against ever-more numerous—and increasingly capable—threats and adversaries.

What is Cybersecurity Risk Management?

This subspecialty of enterprise risk management describes a systematic approach to identifying, analyzing, evaluating, and addressing cyber threats to an organization’s assets and operations. At its core, it involves a continuous cycle of risk assessment, risk decision-making, and the implementation of risk controls intended to minimize the negative impact of cyber incidents.

A proactive cyber risk mitigation approach helps organizations protect critical digital assets and bolster business continuity, legal compliance, and customer trust. By integrating risk management with the organization’s overall strategic planning, cybersecurity teams can prioritize resources efficiently and align their efforts with the business’s risk appetite and objectives.

Why Has Cyber Risk Management Become So Critical?

Getting control over cyber risk is quickly becoming a core requirement for businesses operating in today’s digital ubiquity. The proliferation of digital information and internet connectivity have paved the way for sophisticated cyber threats that can penetrate many of our most robust defenses. With the digital footprint of businesses expanding exponentially, the potential for data breaches, ransomware attacks, and other forms of cybercrime has escalated dramatically.

These incidents can result in devastating financial losses, legal repercussions, and irreparable damage to an organization’s reputation. Furthermore, as regulatory frameworks around data protection become more stringent, failure to comply can lead to significant penalties. Given these conditions, an aggressive and comprehensive approach to managing cybersecurity risks is crucial for safeguarding an organization’s assets, ensuring operational continuity, and maintaining trust with customers and stakeholders.

Effective Cyber Risk Management: A Framework-Based Approach

Adopting a structured, framework-based approach to cybersecurity risk management lets security teams corral the complexity of digital environments with a methodical, strategic mitigation methodology. For most enterprise applications, there’s no need to reinvent the wheel. There are a myriad of established frameworks that can be modified and customized for effective use in nearly any environment.

Perhaps the best known is the National Institute of Standards and Technology (NIST) Risk Management Framework (RMF), a companion to NIST’s well-tested and widely implemented Cybersecurity Framework (CSF). The NIST RMF offers a structured and systematic approach for integrating security, privacy, and risk management processes into an organization’s system development life cycle.

Such frameworks provide a comprehensive set of guidelines that help identify and assess cyber threats and facilitate the development of effective strategies to mitigate these risks. By standardizing cybersecurity practices, organizations can ensure a consistent and disciplined application of security measures across all departments and operations.

This coherence and uniformity are crucial for effectively addressing vulnerabilities and responding to incidents promptly. Equally important, frameworks incorporate best practices and benchmarks that help guide organizations toward achieving compliance with regulatory requirements, thus minimizing legal risks and enhancing the safeguarding of customer data. In essence, a framework-based approach offers a clear roadmap for managing cyber risk in a way that’s aligned with organizational strategic objectives and industry standards.

What follows is a checklist based on the 7-step RMF process. This is just a starting point. A framework to-do list like this can and should be tweaked to aid in reducing and managing specific cyber risks in your unique enterprise environment.

1. Preparation

In this initial phase, organizations focus on establishing the context and priorities for the Risk Management Framework process. This involves identifying critical assets, defining the boundaries, and codifying a risk management strategy that aligns with the organization’s objectives and resources. This is the foundation upon which a tailored approach to managing cybersecurity risk will ultimately be built throughout the system’s lifecycle.

  • Establish the context for risk management and create a risk management strategy.
  • Define roles and responsibilities across the organization.
  • Develop a taxonomy for categorizing information and information systems.
  • Determine the legal, regulatory, and contractual obligations.
  • Prepare an inventory of system elements, including software and hardware.

2. Systems Categorization

Expanding on the categorization step (above), this phase involves identifying the types of information processed, stored, and transmitted to determine potential impact as measured against the information security CIA triad (confidentiality, integrity, and availability). Organizations can assign appropriate security categories to their systems by leveraging a categorization standard such as the Federal Information Processing Standard (FIPS) 199, ensuring that the protective measures taken are tailored to the specific needs and risks associated with the information being handled. This step is crucial as it lays the groundwork for selecting suitable security controls in the later stages of the risk management process.

  • Identify the types of information processed, stored, and transmitted by the system.
  • Assess the potential impact of loss of Confidentiality, Integrity, and Availability (CIA) associated with each type.
  • Document findings in a formal security categorization statement.

3. Selecting Appropriate Security Controls

This critical step begins the safeguarding of information systems against potential threats and vulnerabilities in earnest. Based on the categorization of the information system, organizations select a baseline of security and privacy controls (NIST Special Publication 800-53 or some equivalent controls standard is a good starting point here), corresponding to the system’s impact level. This baseline acts as the jumping-off point for the security controls, which can be tailored to address the specific risks identified throughout the risk assessment process. Customization involves adding, removing, or modifying controls to ensure a robust defense tailored to the unique requirements and challenges of the organization.

  • Select an appropriate baseline of security controls (NIST SP 800-53 or equivalent).
  • Tailor the baseline controls to address specific organizational needs and identified risks.
  • Document the selected security controls in the system security plan.
  • Develop a strategy for continuously monitoring and maintaining the effectiveness of security controls.

4. Implementing the Selected Controls

Implementing security controls involves the physical and technical application of measures chosen during the previous selection phase. This step requires careful execution to ensure all controls are integrated effectively within the environment, aligning with its architecture and operational practices. Documenting the implementation details is crucial to provide a reference for future assessments and maintenance activities.

  • Implement the security controls as documented in Step 3.
  • Document the security controls and the responsible entities in place.
  • Test thoroughly to ensure compatibility and uninterrupted functionality.
  • Prepare for security assessment by documenting the implementation details.

5. Assessing Controls Performance

Assessing security controls involves evaluating effectiveness and adherence to the security requirements outlined in the overall security plan. This phase is critical for identifying any control deficiencies or weaknesses that could leave the information system vulnerable. Independent reviewers or auditors typically conduct assessments to ensure objectivity and a comprehensive analysis.

  • Develop and implement a plan to assess the security controls.
  • Perform security control assessments as per the plan.
  • Prepare a Security Assessment Report (SAR) detailing the effectiveness of the controls.
  • Determine if additional controls are needed and append the master security plan accordingly.

6. Authorizing the Risk Management Program

The authorization phase is a vital decision-making interval where one or more senior executives evaluate the security controls’ assessment results and decide whether the remaining risks to the information systems are acceptable to the organization. Upon acceptance, authorization is granted to operate the mitigation program for a specific time period, during which its compliance and security posture are continuously monitored. This authorization is formalized through the issuance of what is known as an Authorization to Operate (ATO) in some organizations, particularly in the public sector.

  • Compile the required authorization package, including the master plan, the SAR, and the so-called Plan of Action and Milestones (POA&M).
  • Assess the residual risk against the organizational risk tolerance.
  • Document the authorization decision in an Authorization Decision Document.

7. Monitoring and Measuring Against Performance Metrics

The monitoring phase ensures that all implemented security controls remain effective and compliant over time. Continuous surveillance, reporting, and analysis can promptly address any identified vulnerabilities or changes in the operational environment. This ongoing process supports the kind of flexible, adaptive security posture necessary for dealing with evolving threats while steadfastly maintaining the integrity and availability of the information system.

  • Implement the plan for ongoing monitoring of security controls.
  • Report the system’s security state to designated leaders in the organization.
  • Perform ongoing risk assessments and response actions, updating documentation as necessary.
  • Conduct reviews and updates regularly, in accordance with the organizational timelines, or as significant changes occur.

Conclusion: Formalizing Cyber Risk Mitigation

A solid risk management framework provides a comprehensive guide for enhancing the security and resilience of information systems through a structured process of identifying, implementing, and monitoring security controls.

Sticking to a framework checklist helps ensure a successful, systematic adoption. As noted throughout, engaging stakeholders from across the organization, including IT, security, operations, and compliance, is critical to ensuring a truly comprehensive risk management program. Additionally, periodic training and awareness for team members involved in various phases of the risk management project will contribute to the resilience and security of the organization’s digital assets.

Organizations can effectively safeguard their digital assets and mitigate unacceptable risks by following the outlined steps, tailoring the program to fit specific organizational needs, involving stakeholders, conducting regular training, and adapting to the evolving cybersecurity landscape. Ultimately, this kind of formal, structured cyber risk management fosters a culture of continuous improvement and vigilance in an enterprise, contributing to the overall security posture and the success of the organization.

-- Gunter Ollmann

First Published: IOActive Blog - August 20, 2024

Friday, March 1, 2024

GenAI Influencing the Secure Code Review Profession

 It’s tough to be a secure code reviewer. There are already over 700 programming languages according to Wikipedia, and seemingly more languages materializing every year. Expectations are high that rapid developments in Generative Artificial Intelligence (GenAI) will bring a new suite of languages and security issues that’ll have an oversized impact on software development. Consequently, secure software development lifecycle (SDL) processes and security code review are having to evolve rapidly.


I’m both excited and nervous about GenAI advancements in the world of software development and secure application design. It’s exciting to see how prompt engineering of Large Language Models (LLM) and adoption of AI augmentation in the form of copilots and chatbots are increasing the pace of ideation into new products. I’m nervous about the hallucinations and code quality being generated in response though.

English as a Programming Language

2023 was the breakthrough year for AI, with LLM and GenAI permeating every industry, technology, and product. Today, the most in-demand languages currently are Python, C, and C++ but, controversially, the future star programming language may in fact be English; something that’ll take some time to adjust to.

For over a decade we’ve been told that the supply of experienced cybersecurity professionals has trailed the market’s requirements, with a deficit growing year-on-year, and a casual scan across office desks and cubicles will highlight a more significant gender gap across the cybersecurity (and software development) industry. I think AGI and emergence of English as a critical programming language are fundamental to correcting both industry problems.

GenAI, particularly those based upon LLM advancements, are increasingly sophisticated language machines – and women may have an advantage over men in maximizing utility and productivity from them.

Multiple studies over the last 30 years have constantly highlighted that women are better communicators than men. “Better” is obviously an explosive and controversial term even amongst the academics who published the studies, but in general women have more expansive vocabularies and stronger interpretative communication skills. Modern neuroscience and studies in children and adolescents identify girls as more garrulous than boys, with greater complexity and sophistication of language, and tend to develop more in the realm of listening with greater focus and concentration as they age. This historically translates into women being better coders than men (once you remove the bias in the system).

As I look to GenAI and the expanding world of prompt engineering, I anticipate that women will have an advantage over their male developer counterparts. Strong and well-developed communication skills (and the reasoning and understanding that underlays those polished skills) are prerequisites for maximizing efficiency of GenAI-returned results and tuning responses – both now and for the immediate future.

Starter-job Experience

But what about experience? The “experience gap” is often called out as a chasm for newly minted degree-holding graduates and landing a starter-job in cybersecurity.

It’s rare to find an entry-level job in our industry that doesn’t require multiple years of hands-on security experience nowadays as many of those traditional starter roles – network scanning, alert triage, playbook maintenance, patch management – have been automated away, with many more projected to disappear as AI adoption increases.

Most successful new entrants into the cybersecurity profession come from adjacent technical industries making a career jump rather than direct from a college or university. Armed with transferable skills and technical experience, they’re capable of crossing the chasm left in the wake of cyber automation. However, the security knowledge gap between a cybersecurity veteran and a recent transfer remains large and a growing concern for the industry.

I’m excited to think AI augmentation and copilot technologies will have one of the largest impacts on our industry – removing much of the security knowledge gap and reducing the overall impact of the experience gap – like what is happening in other industries, such as the medical field. For example, AI use in patient triage, predictive analytics, and virtual assistants are augmenting generalist regional nurses (two-year qualification) and Bachelor of Science in Nursing (four-year qualification) graduates, and allowing them to perform many of the roles and responsibilities traditionally associated with a completed medical doctor degree (10 to 12 years).

Secure Code Reviews

It’s tough to be a secure code reviewer. There aren’t enough of them. The job requires tremendous amounts of experience and advanced security knowledge, and it’s tiring and hard work.

GenAI is going to have a huge impact on their job.

On the positive side, English as a programming language and AI augmentation and copilots is going to help increase both the breadth and depth of the cybersecurity talent pool available to perform this critical job. The tools available to code reviewers to automatically review and assess code security are advancing quickly and, while still in their first generation of AI adoption, are anticipated to mature rapidly and identify vulnerabilities and logic flaws with higher fidelity and trust. I’m sure there’ll be a gap between the best that a tool can achieve versus the best-of-the-best human expert though – especially when that expert is augmented and using similar tools themselves.

Meanwhile, GenAI is spearheading prompt engineering of new software applications. A new generation of product developers may have little to no influence over the code that powers the application. Indeed, I’ve previously argued that the role of product manager will change greatly in the coming years as their skills in product design and requirement setting pivot from being directed to engineering teams and into GenAI prompts instead.

What does an GenAI-generated application look like under the covers? Time will tell. We anticipate it’ll increasingly become more secure – using best security practices and recycling pretested code behind the scenes – and that it’ll constantly learn, optimize, and apply best and better security practices – but we’ll still need those human secure code reviewers for some time to come, specially when it comes to high-impact applications and certification.

A concern though as GenAI does more of the application development software developers will have less direct influence over the underlying code is that powering the application and business logic. It would be a nightmare if the AGI produced entirely different code throughout the application each time it received new design criteria and re-optimized – making vulnerability triage, reporting, reconciliation, and tracking near impossible, and code reviews and code certifications (human or tool led) largely nonsensical.

Watch this space!

I think we’re still some years away from having to worry about continuously reimagined code generated by GenAI without human software developers tweaking and refining the underlaying application code, but it is tremendously exciting to see the rapid advances in prompt engineering and how LLM’s are being incorporated into both old and new products.

Our industry has consistently struggled to attract and retain women. GenAI has the potential to not only level the field and make it easier to join the game, but to also leverage previously poorly-tapped communication skills for the betterment of both application development and security. There’s a lot of work ahead. There’s a lot of research to be done. There’s a lot of opportunities to make code more secure!

-- Gunter Ollmann

First Published: IOActive Blog - March 1, 2024

Tuesday, June 20, 2023

Why Cybersecurity is Critical in MLOps

Larger and more sophisticated businesses will lean into building out their in-house data science teams and capabilities

If your business relies on machine learning (ML) to drive strategic decision-making, you’re in good company. A recent report by ClearML shows the technology clearly entering the mainstream, with 60% of organizations’ ML leaders planning to increase ML investments by more than a quarter in 2023. The same study revealed that 99% of respondents either already have dedicated budgets for ML operations (MLOps) or plan to implement them this year.

But, as MLOps mature, they also carry more risk. According to a recent study by NCC Group, organizations are deploying ML models in more applications without considering security requirements. In a separate survey by Deloitte, nearly two-thirds of AI and ML users describe cybersecurity risks as a significant or extreme threat, but only 39% feel prepared to combat those risks.


MLOps model creation pipelines are vulnerable and easily attacked in three separate ways: by malicious insiders, through software supply chain manipulation, and via compromised systems. If the SolarWinds supply chain attack taught the industry anything, it’s that continuous build processes are both a target for sophisticated adversaries and a blind spot for in-house security operations teams.

In 2023, continuous build processes will continue to be a target for threat actors. As these attacks start to impact enterprises’ bottom lines, they will have to start paying more attention to the cybersecurity side of MLOps.

Here are some ways to make MLOps projects safer and more secure.

Secure the Whole Pipeline

Part of the challenge of securing MLOps is the sheer length and depth of typical machine learning pipelines. They include a half dozen or more phases – data collection and preparation, along with the creation, evaluation, optimization, deployment and usage of an ML model. Vulnerabilities can crop up at any point in the process.

Early on, in data collection, threat actors can taint the data, manipulate the annotation or conduct adversarial attacks on the metadata stores. In later phases, open-source models and frameworks can include hidden vulnerabilities. Potential bias and system performance need to be addressed. And as models are deployed and used, new data is often introduced, expanding the attack surface and opening an organization up to all kinds of threats – including evasion attacks, model theft, code injections and privacy attacks.

At the tail end of the process, there’s a lot of intellectual property inside an ML model. Decades of transactional data and learnings from financial models that are built and trained into models may only be 10s of kilobytes in size. It’s easier to steal that model than to steal the actual source data.

These models tend to be exposed. Attackers have become skillful at querying models and reproducing them somewhere else. This requires a new way of thinking about the value of the model. Tooling and alerting not only around the theft of data but around the manipulation of the models is important to an overall MLOps security strategy.

Invest in Tooling to Scale Across SOCs

It’s no secret that security is no longer siloed in a single department. It cuts across all functions, and organizations are creating Security Operations Centers (SOCs) to improve the visibility, manageability and auditing of their overall security posture. To extend the SOC’s capability to MLOps, organizations need to incorporate tooling that scales to much larger uses than ever before.

Meeting MLOps’ data needs forces SOCs to adapt in two ways. Existing SOC operations teams are now accountable, forcing them to build the additional tooling and reporting to support MLOps teams from a security perspective. Plus, MLOps teams that are specialized in data science curation are able to leverage larger toolsets – including logging analytics platforms that provide higher levels of threat detection.

Double Down on Security Best Practices

Some of the best defense tactics for MLOps are practices organizations deploy regularly across the rest of their operations. A zero trust security policy requires the authentication and authorization of anybody trying to access applications or data used in the development of ML models. It also tracks their activity. Applying the principle of least privilege (PLoP) limits users’ access to the exact data sets and models they are authorized to touch. This reduces the attack surface by prohibiting hackers who have gained access to one data trove from moving freely throughout the system.

Use Analytics to Observe and Log ML Tasks

An important step in protecting an ML system is to understand the system’s behavior in healthy and unhealthy states. To do this, organizations need to set up alerts that trigger action before an incident occurs. This is called “observability.” A vulnerability introduced early in the training data will affect the model’s performance down the line. Tracking performance data and logging metrics of ML tasks gives organizations insights into any and all security issues that could affect the ML model.

Future Monitoring of Model Development Lifecycle

The continuous lifecycle of MLOps necessitates the continuous monitoring of a deployed model’s response to adversarial manipulation and corruption. In the future, expect to see larger and more sophisticated businesses lean into building out their in-house data science teams and capabilities, detecting threats with security analytics, and pruning and filtering data from unknowns to improve the advancements in the next generation of ML and AI.

-- Gunter Ollmann

First Published: Security InfoWatch - June 20, 2023

Tuesday, March 29, 2022

Why the SOC Needs to Learn from the Aviation Industry

The cybersecurity industry has spent a lot of time talking about improving the analyst experience while not making significant improvements, as much of the efforts have been too focused on finding a silver bullet solution. Combine that with a global pandemic and now things are just getting worse. A recent study published by Devo, the 2021 SOC Performance Report, found that on a 10-point scale, where 10 indicates SOC staff have a “very painful” experience performing their jobs, 72% of respondents rated the pain of SOC analysts at a 7 or above.


Instead of thinking about the aforementioned silver bullet to alleviating SOC pain, I wanted to focus on one of the top sources, alert fatigue, and how the cybersecurity industry might be able to take a page out of another field to find a solution.

In the SOC Performance Report, a whopping 61% said a cause of SOC pain was that there are too many alerts to chase. I think it’s safe to draw the connection that “alert fatigue” will expand to “posture fatigue” and “policy fatigue,” as it adversely affects both recruitment and all too critical retention of experienced SOC professionals.

Alert fatigue may exit the aircraft

So, if we can’t figure out within the security industry, let’s learn from others. There are many non-cyber industries and professions that suffer similarly with alert fatigue, and perhaps the cybersecurity industry can reapply some of those learnings. Across these compatriots of alert fatigue, if we ask the question “how do alarms, warnings, and alerts differ?” I think we’ll find much similarity and overlap in answers — in both the theory and practice of how human operators are supposed to respond and how they do so in reality.

For the purpose of this article, I want to take a look at the aviation industry as our example to the SOC. They have navigated many of the problems SOC operators face today and have made the most progress in governing and managing the ergonomics of sensory overload and automation. Picture this: the inside of an airplane cockpit with all its knobs, buttons, lights, and alerts isn’t too dissimilar to the combined dashboards SOC analysts have to navigate when triaging, investigating, and responding to threats.

In 1988, The Washington Post reported on a “glass cockpit” syndrome in the aviation industry, that reads eerily similar to what many say or think about the SOC today. Researchers from the American Psychological Association noted that pilots would “fall victim to information overload and ignore the many bits of data pouring from myriad technical systems,” and that in airline crashes they studied it was found that “black box recordings showed that the crews talked about ‘how the systems sure were screwed up’ but did not verify what was wrong. In both cases, the systems worked but crews failed to check the information and crashed.”

Similarly, research published in 2001 by the Royal Institute of Technology examined “the alarm problem” in aviation, meaning, “in the most critical situations with the highest cognitive load for the pilots, the technology lets you down.” The reports noted that “the warning system of the modern cockpits are not always easy to use and understand. The tendency is to overload the display with warnings, cautions and inoperative system information accompanied by various audio warnings.” It went on to identify one of the main problems as a result of this overload as “a cognitive problem of understanding and evaluating from the displayed information which is the original fault and which are the consecutive faults.” Sound familiar? You would likely hear something extremely similar from someone working in today’s SOC.

In the decades that followed, aircraft cockpit design has progressively applied new learnings and automation to dynamically manage alert volume and the attention of the pilot to priorities. In the Royal Institute of Technology’s report, researchers identified accident simulation as an effective tool for improving cockpit alert systems, finding more associable ways to present alerts such as differentiating sounds and the introduction of context, which would allow pilots to “immediately understand what part or function of the aircraft is suffering a malfunction.” More context would also include guidance on what to do next. In its conclusion the study noted:

Such simulations would hopefully result in less cognitive stress on behalf of the pilots: they would know that they have started to solve the right problem. They would not have to worry that they have entered the checklist at the wrong place. With a less stressful situation even during malfunctions there is greater hope for correct actions being taken, leading to increased flight safety.

SOC systems need to embrace and apply many of these same learnings that have spanned decades for aviation. The majority of the cybersecurity industry seems to have only gotten as far as color coding alert and warning significance, leaving the analyst faced with a hundred flashing red priorities, even after triaging it. It’s no surprise that they’re both overwhelmed and unable to respond to complex threats across a broadening attack surface.

Beware of Autopilot

When it comes to solving the issue of alert fatigue, automation is typically one of the first things to come to mind. The same went for aviation in 1988, where the previously mentioned Washington Post report quoted researchers saying what could have been taken right from a security trade publication in 2022:

Research is badly needed to understand just how much automation to introduce — and when to introduce it — in situations where the ultimate control and responsibility must rest with human operators, said psychologist Richard Pew, manager of the experimental psychology department at BBN Systems and Technologies Corp. in Cambridge, Mass.

“Everywhere we look we see the increasing use of technology,” Pew said. “In those situations where the operator has to remain in control, I think that we have to be very careful about how much automation we add.”

The growing use of high-tech devices in the cockpit or on ships can have two seemingly contradictory effects. One response is to lull crew members into a false sense of security. They “regard the computer’s recommendation as more authoritative than is warranted,” Pew said. “They tend to rely on the system and take a less active role in control.” Sometimes crews are so mesmerized by technological hardware that they are lulled into what University of Texas psychologist Robert Helmreich calls “automation complacency.”

And while automation of course has an important part to play in incident response and investigation — just as it does in modern aircraft cockpit design — it comes with some key warnings:

  1. Situational awareness is lost. Automation is often brittle, unable to operate outside of the situations it is programmed for, and subject to inappropriate performance due to faulty sensors or limited knowledge about a situation.
  2. Automation creates high workload spikes (such as when routine changes or a problem occurs) and long periods of boredom (in which attention wavers and response to exceptions may be missed). If you’re staffing for automation-level activities, how do you manage capacity for spikes?

The SOC Earns its Wings

As an industry we have to take a page from the aircraft handbook and avoid increasing cognitive demands, workload and distractions, and make tasks easier to perform. But we must also understand how to manage automation failure and exceptions better.

  • Embrace AI and autocomplete: Like the more advanced sentence autocomplete functions appearing in email and word processing applications, SOC analysts are still in charge of managing an incident, but there is an opportunity to further guide and preemptively enrich a threat investigation, thereby increasing the speed and robustness of response.
  • Distill and prioritize at the incident level, not the alert level: It’s not about filtering/correlating/aggregating alerts, it’s about contextualizing both events and alerts in the background and only articulating an incident in plain single-sentence language. Analysts can double-click down from there.
  • Leverage a community of experts: As attack surfaces increase and vertical technology specialization becomes tougher for in-house SOCs to cover (particularly in times of competing incident prioritization), it becomes increasingly important to be able to “phone-a-friend” and access an on-demand global pool of expert talent. It’s like having several Boeing engineers sitting in the cockpit with the pilot to troubleshoot a problem with the plane.

-- Gunter Ollmann

First Published: Medium - March 29, 2022

Thursday, December 9, 2021

You May Not Have Asked, But The SOC Evolution Answered Anyways

Let’s get the obvious out of the way: The attack surface is growing exponentially and diversely.

Bigger shark, same small boat

The environments, platforms, services, regions and time zones that constitute modern enterprise operations and drive digital transformation for business continue to require increasing specialization and expertise beyond current in-house capabilities. Through a security lens, enterprise attack surfaces are expanding beyond the business’ ability to protect.

Meanwhile, global hiring and retention of security expertise continues to be a weak spot, and direct access to specialized security knowledge and experience is becoming increasingly difficult and costly. And while all that is going on, the volume, duration, pace and sophistication of attacks continues to increase and require significant acceleration in SOC response times and durability — and subsequent autonomous response systems.

Saying we’re in a conundrum is vastly understating things.

The security industry is at the gate of a forced SOC evolution, and as you can see, pressure is coming from all directions to drive that change.

The more things change, the more they stay the same

Plenty has happened that has tried to look like an evolution. For the last decade the security industry that powers SOCs has fixated on automation as the key to alleviating some of the pressures. But after a decade, have things really changed?

SOAR was a brief shining light that has come and mostly gone, having been absorbed back into SIEM, as the legacy SIEM vendors acquired dedicated SOAR vendors to make up for their shortcomings in human workflow automation. This didn’t solve much, as analysts were more or less left in the lurch. They faced the same automation integration challenges, only now they’re locked into a single vendor (where previously an “independent” SOAR offered the prospect of multi-vendor connectors and flexibility to operate independently of SIEM lock-in).

And that’s not the end of our automation woes, either. Automation, on its best day, is still too playbook-oriented. To get things done, experts have to essentially write scripts for each new system, connector and application in an enterprise. If we had set out to create librarians out of analysts, that’s an area our industry could say it had actually achieved success in.

But in all seriousness, we’re caught in a linear script development cycle and automation hasn’t yielded the reduction in analyst workloads that we so desperately need.

I’d like to get off the ride, please

So how do we break the cycle? I can identify two major breakthroughs that will move the needle forward for the SOC evolution.

First, the successful implementation and use of AI “smart” orchestration systems within the SOC.

I’m sure many SOC analysts and CISO’s are jaded from past promises, but the reality is that AI and ML approaches have matured significantly over the last year, and have reached the inflection point of their “hockey stick” usefulness trajectory and the value they can bring. I think as an industry it’s time we start to move past our fear of turning on automated response and protection capabilities that are powered by this new generation of AI and ML. By embracing it, SOCs will become much more effective at detection, which will lead to a reduction in the number of distinct alerts and false positives (put that in the win column for reducing analyst workloads).

Second breakthrough: The ability to tap a global community of contributors via marketplace ecosystems, or more simply put, sharing is caring.

Detection-as-code, policy-as-code, blah-as-code has redefined content development and vendor-proprietary product-dependent content. Platform-independent content (ranging from alerts, threat detection, playbooks, etc.) is rapidly and readily available from a global array of sources, and availability will continue to increase. The ability to tap a global pool of expertise is more prevalent than ever and it feels like the gig economy is finally coming to the security world via the SOC. I think this would have surprised many people just a few years ago, but in the wise words of one Jim Carrey — “desperation is a necessary ingredient to learning anything.”

I don’t care how, I want it now

Well, you can’t have it…yet. But you can start. Both “smart” machine-intelligence and content marketplaces directly address the pressure points previously mentioned, but the industry is still in early stages of the SOC evolution. Right now organizations have to take a look at their SOC and decide how they’re going to reorganize and prioritize to discover and implement the people, tools and partners they’ll need to usher in the evolution.

There are some philosophical hurdles to be overcome, but I believe business needs will drive the pace of change. It used to be the case that penetration testing was in-house only, then extended to trusted vendors managed under restrictive agreements, and on to industry-accredited providers, and now businesses can tap broad communities of bug-bounty-based individual contractors and cloud-based automated attack simulators. If we managed those industry changes, I’m pretty sure we can manage the same for incident response and investigation.

-- Gunter Ollmann

First Published: Medium - December 9, 2021

Tuesday, December 22, 2020

Attesting to the Security of Data-in-Use

The pace at which new confidential computing solutions are penetrating enterprise security architectures and data protection strategies appears to be catching security leaders off balance. COVID-19-accelerated digital transformation saw years’ worth of cloud migration, “zero trust” management and online collaboration tool rollouts squeezed into a few short months. Solutions engineering and security teams that thought they’d have a couple years to learn and master the next set of security- and privacy-preserving technologies are suddenly playing catch-up in the newly “cloudified” enterprise.

Having already mastered and commodified “data-at-rest” and “data-in-transit” security, security leaders are under pressure to support companies’ adoption of confidential computing technologies and newly enabled trusted execution enclave (TEE) services. If 2020 represented a step function for digital transformation and cloud adoption for businesses, 2021 will be the year of rapid, measurable “data-in-use” security and privacy. 

As the list of new and pending TEE-enabled products and services from major public cloud providers grows, where should CISOs and security architects begin? For most organizations, the two most influential confidential computing building blocks will be enclave attestation and enclave-enabled relational databases.

Whether the organization is planning on utilizing in-house or cloud computing builds atop Intel SGX or AMD SEV chip architectures (or Arm, NVIDA, etc. in the future), attestation lays at the heart of confidential computing trust. Enclave provisioning and trust will quickly become as fundamental to enterprise security as identity management, certificate management and key management.


Enclave attestation services are designed to verify and validate that the confidential computing workload is provisioned and executed securely in a TEE environment. Remote attestation services will necessarily vary in environmental specifics, but they generally validate the integrity — the root of trust, hardware status, firmware status, security patch status, etc. — of the TEE (hardware or virtualized) before releasing sensitive data into the enclave. It is therefore the attestation service’s responsibility to cryptographically ensure that the underlying hardware and firmware is not in a vulnerable state before use, that the workload assigned to the enclave is transferred securely and, upon execution, that the enclave remains secure and the workload’s computing functions and output were not tampered with in any way. In essence, confidential computing attestation tells you if the results that your code generated within the TEE are trustworthy.

It won’t be long before CISSP study materials include attestation mechanics with an “Alice and Bob” explanation, like those used to teach data flows that constitute asymmetric cryptography and key exchange mechanics. For most InfoSec professionals, a basic understanding of enclave attestation will be adequate for traffic routing and troubleshooting. A deeper understanding will be required for architects and DevOps teams tasked with deploying and trusting new confidential compute workloads.

In parallel to incorporating enclave attestation service design into new business application architectures, security and privacy leaders will need to leverage enclave-enabled relational databases — especially if they’re to meet toughening regulatory requirements linked to customer data privacy. 

In the database world, balancing privacy and “data-at-rest” encryption with data utility and business application performance has been a delicate, often compromising, affair. Encrypted column data (for example, customer names, addresses and blood types) logically makes it more difficult to perform searches and match records. The use of deterministic encryption techniques (for instance, encryption that always generates the same encrypted value for any plain text value) is vulnerable to some predictive attack vectors. However, randomized encryption techniques are more secure but make most common query types impossible.

To protect sensitive data from malware and high privileged unauthorized users of the database server, traditional non-TEE data encryption processes protect the data by encrypting it on the client side. This means disallowing the data or corresponding cryptographic keys from appearing in plaintext inside the database engine. When deterministic encryption is used, the only operations the database engine can perform are equity comparisons. All other operations, including cryptographic operations (initial data encryption or key rotation) or rich computations (for example, pattern matching), are not supported inside the database. Users need to move their data outside the database to perform these operations on the client side. These actions are completed as a secure software agent to transport encrypted column data to a remote host or system to decrypt and perform query actions upon the data, then re-encrypt data that needs to be written back to the protected column.

For operations teams tasked with regulatory data discovery, labeling and protection throughout the enterprise, the mechanics of securing client agents and shuffling encrypted data between systems —temporarily duplicating data in the process — is inefficient and burdensome. TEE-enabled database services ensure the encrypted data remains within the system, allowing computations on plaintext data inside the secure enclave with no way to view data or code inside the enclave from the outside (even with a debugger). In addition, rich computations, such as operations on encrypted columns, are possible, and cryptographic operations on sensitive data, like initial data encryption or rotating a column encryption key, are performed within the enclave and do not require moving the data outside the database.

As confidential compute services become ubiquitous, enclave attestation and enclave-enabled relational database technologies will be fundamental building blocks for post-COVID-19 business application design and delivery. CISOs and their security teams need to quickly master these technologies if they’re to successfully partner with in-house development teams and secure “data-in-use.”

-- Gunter Ollmann

First Published: SecurityWeek - December 22, 2020

Tuesday, November 3, 2020

Securing Data-in-Use With Confidential Computing

Confidential Computing Capabilities and Services Will be a Competitive Differentiator for Cloud Providers

Most organizations should have mastered by now the first two pillars of data protection – securing data-at-rest and data-in-transit – across their hybrid enterprise environments. The third data protection pillar – securing data-in-use (i.e., protecting and encrypting data in use while in memory and during computation) – has been elusive, but is in the process of being addressed through the transformational motion commonly referred to as confidential computing. 

Technology business leaders are ideally pursuing transformation plans that assume ubiquitous confidential computing availability and data-in-use security will be a cloud-native default within five years.

For many organizations, completing their digital transformation journey has been conditional on being able to categorically ensure that absolutely no one – not a trusted system administrator, the OS developer, the cloud provider, law enforcement, malicious insiders, or an attacker armed with powerful zero-day exploits – can ever secretly access or manipulate the data and intellectual property they entrust to the cloud. Consequently, as the third pillar of data security, confidential computing will increasingly be a prerequisite for any cloud-deployed business application.


The technologies, platforms, and architectures that enable confidential computing have evolved at an astounding pace – especially when compared with the decades it has taken for data-at-rest encryption to evolve from password-protected ZIP files in the early 1990s to today’s enabled-by-default hardware-based encryption locked to the physical compute system, or the continued effort to transition data-in-transit defaults from HTTP to secure HTTPS (preferably using TLS v1.3).

The global pandemic has not held back public cloud advancements and new service offerings in confidential computing. Virtualization infrastructure for confidential computing built atop hardware-based trusted execution environments (TEEs) on servers that implement Intel Software Guard Extensions (Intel SGX) are generally available along with previews of confidential VMs using hardware-based TEE on servers supporting AMD’s secure encrypted virtualization (AMD SEV) extension. In parallel, confidential computing options have begun extending across cloud services to embrace Kubernetes confidential nodes, always encrypted SQL databases, confidential machine learning interfaces, HSM key management, and IoT edge compute.

It can be difficult for security leaders to keep pace with underlying hardware advances and their applicability. For example, the memory integrity protection features of Intel SGX are well suited for highly security-sensitive but small workloads, while AMD SEV is useful for “lift and shift” of existing complex or legacy applications and services without necessarily refactoring existing code. Meanwhile, Intel’s trust domain extensions (Intel TDX) will enable hardware isolated virtual machines (called trust domains), AMD’s secure encrypted virtualization-encrypted state (SEV-ES) will help ensure that guest VMs are encrypted and attestable once they stop running, and many more hardware advances from Intel, AMD, Arm, NVIDIA, etc. that help mitigate new and potentially intrusive memory, compute, and attestation threats will become available via the major cloud providers in the coming year or two.

Clearly, confidential computing is in a transitional period as these new hardware-enabled solutions get adopted and deployed by major cloud providers and their most advanced customers. 

While it is easy to get lost in the hardware-based security feature advances the silicon providers are delivering, the long-term assumption security leaders should be planning for is that the physical infrastructure will verifiably guarantee that enclaved processes, memory, and the data they hold or manipulate will be secure from any and all prying eyes – in particular, the cloud and software stack providers – and that all cloud services (from leading public cloud providers) will operate in a secure data-in-use mode by default. It is reasonable to assume that within five years the term “confidential compute” will become superfluous and an assumed native component of all cloud services.

In the meantime, confidential computing capabilities and services will be a competitive differentiator for the large cloud providers. 

As the underlying hardware advances and public cloud providers extend data-in-use security, integrity, and attestation capabilities across their customer-available services, business technology leaders will need to assess each cloud service individually and assume some short period of cloud-specific lock-in for the custom applications their own business will engineer.

Opportunities ranging from anti-money-laundering to customer analytics within the financial services industries, privacy-preserving collaborative disease diagnostics and drug development within healthcare services, joint intelligence analysis and anti-corruption across government agencies, etc., are a snippet of newly achievable privacy and confidentiality-preserving solutions from organizations adopting cutting edge confidential computing-enabled cloud solutions.

Some enterprises may be content to wait until data-in-use security becomes ubiquitous to complete their digital transformation. While confidential compute capabilities are expanding and new secure cloud services are “fluid” in their evolution, there is a clear window for technology-agile businesses to innovate and partner with their cloud provider to bring new categories of secure products to market ahead of both competitors and regulators.

-- Gunter Ollmann

First Published: SecurityWeek - November 3, 2020

Tuesday, June 9, 2020

Navigating the Rapid Digital Shift: Ticket on the Bus, Not the Whole Bus

Global Companies’ Evaluation of Cybersecurity Solutions Selection Has Been Steadily Changing 

If it wasn’t already obvious to cybersecurity sales teams, there’s been a sea change for large organizations evaluating and buying new security products to protect their businesses. Responding to COVID-19, transformation plans that enable “work from home” such as Zero Trust identity and access management have been greatly accelerated, while technology refreshes and other capital-intensive plans are being pushed back.

Now, several months into this new operations paradigm, there may be added credence to the adage “in for a penny, in for a pound.” 


Many large companies have successfully navigated the digital shift to most of their workforce working remotely, finding the transition less difficult than first envisaged and achieving higher productivity than anticipated. Because such companies have resolved long-held internal conflicts over the security and integrity of cloud-based business operations, many of those postponed capital-intensive projects are being reviewed with a cloud-enabled, subscription-based lens.

This has several ramifications for cybersecurity vendors—particularly the specialized boutiques and innovative startups looking to quickly capitalize on new security opportunities.

Global companies’ evaluation of cybersecurity solutions selection has been steadily changing over the past couple years. The rapid digital shift of recent months has reinforced the need for change. 

I’d like to offer advice to vendors attempting to reach out and position their new cybersecurity products.

  1. “I’ll buy a ticket, not the whole bus.” For decades, startups have looked to the largest companies as the Golden Goose and focused great energies in selling into them. The premise being that by solving a critical problem for them at a very high premium, that will cover the costs of developing an actual solution that can be sold broadly—e.g., the sale will fund my company’s product development. Although there may be a few cases where only a custom-tuned solution is required, many large businesses now prefer to buy a close-enough solution off the rack and work with the vendor as an advisor—not an investor. CISOs are looking at the sustainable list price of the solution and will purchase at a discounted level proportional to their deployment’s scale.
  2. “Cost projection is critical.” Although highly versatile and scalable, cloud-based services billing can be difficult to predict—especially if the cybersecurity solution requires multiple third-party and cloud-provider SaaS dependencies. Security owners and budget holders are requiring vendors to provide accurate billing forecast and tiered discount models for the complete solution—models that include all dependent service costs (e.g., log storage analytics, container management). Vendors need to remove as much calculus from the pricing as possible and be prepared for billed services to be pared back if overly optimistic projections exceed the planned budget. Cost discussions have replaced those about cloud solutions prices.
  3. “Features must be pre-integrated.” If the product is a feature (which, let’s face it, almost all new startup products are!), recognize it as a feature and don’t position it as a partial solution. As a feature product, integration with the solutions businesses already use is a prerequisite, and sales representatives should lead with the integration and interoperability first. CISOs are looking to shrink their attack surface and simplify the portfolio of products and vendors they rely on, and are increasingly reluctant to take on the task of brokering partnership between vendors as a prerequisite for extracting new protection value. Feature products benefit greatly by being enabled from within a solution provider’s product or marketplace.

On a related note, with the surge to execute day-to-day business operations remotely with a diverse and globally distributed workforce, cybersecurity buying decisions will increasingly factor accessibility, usability, and inclusiveness in solution design and operability. Vendors will be steered toward cloud-standardized accessibility interfaces—enabling visually impaired employees to use screen readers or dexterity-limited users to employ voice-to-text controls—to perform their analysis.

These changes are not unique to the largest enterprise businesses and are trickling down to other educated cybersecurity buyers feeling the same buying pain. Forewarned is forearmed.

-- Gunter Ollmann

First Published: SecurityWeek - June 9, 2020

Thursday, November 14, 2019

Securing Autonomous Vehicles Paves the Way for Smart Cities

As homes, workplaces, and cities digitally transform during our Fourth Industrial Revolution, many of those charged with securing this digital future can find it difficult to “level up” from the endpoints and focus on defining and solving the larger problem sets. It is easy to get bogged down in the myriad of smart and smart-enough devices that constitute “IoT” in isolation of the overall security scope of the smart city – losing both valuable context and constraints.

While “smart city” can mean a bunch of things to different people, for city planners and officials, it’s definition and implementation problems are quite well understood. The vendors that come knocking on their doors promote point solutions – smart traffic control systems, 5G and ultra-high bandwidth wireless communications, driverless vehicles, etc. – leaving the cities’ IT, operational technology (OT), and infosec teams to bring it all together.

An essential part of a security professional’s work is diving deep into the flaws and perils of individual products and clusters of technologies. But trying to “solve security” at a city level is an entirely different paradigm.


A substantial number of my peers and security researchers I’ve worked with over the past couple of decades have focused their energies on securing autonomous vehicles. The threats are varied – ranging from bypassing emission and speed controls to evading the next generation of city road taxes and insurance regulations to malicious remote control of someone else’s vehicle – yet mostly isolated to the vehicles themselves. From what I’m seeing and hearing, they’re doing a great job in securing these vehicles. Their security successes also advance traditional transit solutions, which helps smart cities keep pace with the transportation needs of a growing population. 

Given the continued urbanization of human population, the growth and attraction of megacities (10 million plus inhabitants), and the strains on traditional transport systems, the thought of increasing personal-use autonomous vehicles in these heavily congested cities is outdated and arguably ludicrous. Today’s megacities are already battling traffic congestion with zoned charging, elimination of fossil fuels, and outright banning of private transport. Tomorrow’s megacities – jumping from 33 cities today with the largest holding 38 million people to over 100 with populations in excess of 88 million people by 2100 – need to completely rethink their transport systems and the security that goes with it.

Oddly enough, securing mass transit for megacities come with some advantages. Mass transport systems that evolve from trains, trams, and subways, have embedded within them design constraints that positively influence security. For example, driverless cars of today have to navigate and solve all kinds of road and traffic problems while trams stick to pre-defined paths (i.e. rail networks) with greatly simplified routing and traffic signaling. Research papers covering adversarial AI in recent years have focused on attacking deep learning and cognitive AI systems used by autonomous vehicles (e.g. adding stickers to a stop sign and making the driverless car think the sign says 45 mph), but these tactics would have negligible to no impact on reasonably scoped public transport systems.

It is reasonable to assume that the smart cities of the near future will consist of trillions of smart devices – each of them semi or fully managed, providing alerts, logs, and telemetry of their operations. For those city leaders – particularly CIOs, COOs, CTOs, CISOs, and CSOs – the changes needed to manage, secure, certify, and govern all these devices and their output are mind bogglingly huge.

Interestingly enough, the framework for managing data security for millions of chatty networked devices has largely been solved. Having become cloud-native, modern Security Incident and Event Management (SIEM) technologies have proved to be remarkably successful in identifying anomalies, attacks, and misconfigurations.

The data handling capabilities and scalability of cloud-native SIEM may be just the right kind of toolkit to begin to solve smart city operations (and security) at the megacity level. In addition, with advanced AI being a core component of SIEM, the systems that identify and construct attack kill chains and mitigate threats through conditional access rules could instead be used and trained to identify surge transport requirements (due to concerts ending on a rainy day) and automatically reroute and optimize tram or bus capacity to deliver citizens safely (and dryly) to their destinations – as an example. 

Securing smart cities offers many opportunities to rethink our assumptions on security and “level up” the discussion to solve problems at the ecosystem level. Advancements in AI analytics and automated response technologies can handle the logs, alerts, and streaming telemetry that contribute to OT infrastructure security for mega cities. In turn, this increase in data volume fine tunes anomaly and behavioral-based detection systems to operate with higher efficiency and fidelity, which helps secure city-wide IT infrastructure.

-- Gunter Ollmann

First Published: SecurityWeek - November 14, 2019

Tuesday, April 30, 2019

To Reach SIEM’s Promise, Take a Lesson From World War II

With two of the largest public cloud providers having launched their cloud Security Information and Event Management (SIEM) products and an inevitability that the remainder of the top 5 cloud providers will launch their own permutations some time this year, 2019 is clearly the year of the cloud SIEM.

For an on-premises technology that has been cursed with a couple decades of over-promising, under-achieving, and eye-watering cost escalation, modernizing SIEM into a cloud native security technology is a watershed moment for the InfoSec community.

The promise of finally being able to analyze all the logs, intelligence, and security data of an enterprise in real-time opens the door to many great and obvious things. We can let the SIEM vendors shout about all the obvious defensive value cloud SIEM brings. Instead, I’d like to focus on a less obvious but arguably more valuable long-term contribution that a fully capable cloud SIEM brings to enterprise defense.

Assuming an enterprise invests in bringing all their network logs, system events, flow telemetry, and security events and alerts together into the SIEM, businesses will finally be able to track threats as they propagate in an environment. Most importantly, they’ll be able to easily identify and map the “hotspots” of penetration and compromise, and remedy accordingly.

A unified view will also allow analysts and security professionals to pinpoint the spots where compromises remain hidden from peering eyes. As enterprises strive to deploy and manage an arsenal of threat detection, configuration management, and incident response tools in increasingly dynamic environments, visibility and coverage wax and wane with each employee addition, wireless router hook-up, application installation, or SaaS business connection. Those gaps, whether temporary or permanent, tend to attract an unfair share of compromise and harm.

In World War II, a gentleman by the name of Abraham Wald was a member of Columbia University’s Statistical Research Group (SRG). One problem SRG was tasked with was examining the distribution of damage to returning aircraft and advise on how to minimize bomber losses to enemy fire. A premise of the research was that the areas of bombers that were most damaged and therefore susceptible to flak should be redesigned and made more robust. Wald noted that such a study was biased to only aircrafts that survived their missions and, if you were to assume that damage was more uniformly distributed to all aircrafts, those that returned had actually been hit in the less vulnerable parts. By mapping the damage done to the surviving aircraft, the “undamaged” areas represented the most vulnerable parts of the aircrafts that didn’t survive to return.


Wald’s revelations and work were seminal in the early days of Operational Research – a discipline of applying advanced analytical methods to help make better decisions. I expect cloud SIEM and the integration of AI systems to usher Operational Research and its associated disciplines into the information security sector. Securing an enterprise is a highly complex and dynamic problem and, because Operational Research is focused on optimizing solutions for complex decision-making problems, it is well suited to finding solutions that balance the multi-faceted aspects of business continuity and risk.

As we’re in the early days for cloud SIEM, I’ve yet to see much in the area of employing native AI to address the cold-spots in enterprise threat visibility. The focus to-date is applying AI in threat hunting and automating the reconstruction of kill chain associated with an in-progress attack and supplementing that visualization with related threat intelligence and historical data artifacts.

Putting on a forecasting hat, I expect much of the immediate adoption and growth of cloud SIEM will be driven by desire to realize the promises of on-premises SIEM, in particular, using supervised-learning systems to automate the detection and mitigation of the threats that have pestered security operations teams for twenty-plus years. Infusing SIEM natively on the cloud provider’s platform also creates end to end visibility into security related events inside a business’ environment and pieces in valuable intelligence from the cloud provider’s operations – thereby harnessing the “cloud effects” of collective intelligence and removing the classic requirement for a “patient zero” to initiate an informed response.

What I hope is, once engineering teams have matured those hunting and mitigation capabilities by weaving in AI decision systems and real-time data processing, the “science” of information security can finally come up for air and move forward.

Leveraging the inherent power and scale of public cloud for real-time analytics of enterprise security data at streaming rates means that we’re at the cusp of finally calculating the ROI of each security technology deployed inside an enterprise. That alone should have many CISOs and CFOs jumping for joy. With all the enterprise security data flowing to one place, the cloud SIEM also becomes the anchor for IT operations – such as tracking the “meantime between failures” (MTBF) of protected systems, providing robustness metrics for software assets and system updates, and surfacing the latent risks of the environments being monitored.

75 years may separate War World II from cloud SIEM, but we’re on the cusp of being able to apply the hard-earned learnings from Abraham Wald in our latest adversarial conflict – the cyberwar.

-- Gunter Ollmann

First Published: SecurityWeek - April 30, 2019

Wednesday, March 26, 2014

A Bigger Stick To Reduce Data Breaches

On average I receive a postal letter from a bank or retailer every two months telling me that I’ve become the unfortunate victim of a data theft or that my credit card is being re-issued to prevent against future fraud. When I quiz my friends and colleagues on the topic, it would seem that they too suffer the same fate on a reoccurring schedule. It may not be that surprising to some folks. 2013 saw over 822 million private records exposed according to the folks over at DatalossDB – and that’s just the ones that were disclosed publicly.

It’s clear to me that something is broken and it’s only getting worse. When it comes to the collection of personal data, too many organizations have a finger in the pie and are ill equipped (or prepared) to protect it. In fact I’d question why they’re collecting it in the first place. All too often these organizations – of which I’m supposedly a customer – are collecting personal data about “my experience” doing business with them and are hoping to figure out how to use it to their profit (effectively turning me in to a product). If these corporations were some bloke visiting a psychologist, they’d be diagnosed with a hoarding disorder. For example, consider what criteria the DSM-5 diagnostic manual uses to identify the disorder:

  • Persistent difficulty discarding or parting with possessions, regardless of the value others may attribute to these possessions.
  • This difficulty is due to strong urges to save items and/or distress associated with discarding.
  • The symptoms result in the accumulation of a large number of possessions that fill up and clutter active living areas of the home or workplace to the extent that their intended use is no longer possible.
  • The symptoms cause clinically significant distress or impairment in social, occupational, or other important areas of functioning.
  • The hoarding symptoms are not due to a general medical condition.
  • The hoarding symptoms are not restricted to the symptoms of another mental disorder.

Whether or not the organizations hording personal data know how to profit from it or not, it’s clear that even the biggest of them are increasingly inept at protecting it. The criminals that are pilfering the data certainly know what they’re doing. The gray market for identity laundering has expanded phenomenally since I talked about at Blackhat in 2010.

We can moan all we like about the state of the situation now, but we’ll be crying in the not too distant future when statistically we progress from being a victim to data loss, to being a victim of (unrecoverable) fraud.

The way I see it, there are two core components to dealing with the spiraling problem of data breaches and the disclosure of personal information. We must deal with the “what data are you collecting and why?” questions, and incentivize corporations to take much more care protecting the personal data they’ve been entrusted with.

I feel that the data hording problem can be dealt with fairly easily. At the end of the day it’s about transparency and the ability to “opt out”. If I was to choose a role model for making a sizable fraction of this threat go away, I’d look to the basic component of the UK’s Data Protection Act as being the cornerstone of a solution – especially here in the US. I believe the key components of personal data collection should encompass the following:

  • Any organization that wants to collect personal data must have a clearly identified “Data Protection Officer” who not only is a member of the executive board, but is personally responsible for any legal consequences of personal data abuse or data breaches.
  • Before data can be collected, the details of the data sought for collection, how that data is to be used, how long it would be retained, and who it is going to be used by, must be submitted for review to a government or legal authority. I.e. some third-party entity capable of saying this is acceptable use – a bit like the ethics boards used for medical research etc.
  • The specifics of what data a corporation collects and what they use that data for must be publicly visible. Something similar to the nutrition labels found on packaged foods would likely be appropriate – so the end consumer can rapidly discern how their private data is being used.
  • Any data being acquired must include a date of when it will be automatically deleted and removed.
  • At any time any person can request a copy of any and all personal data held by a company about themselves.
  • At any time any person can request the immediate deletion and removal of all data held by a company about themselves.

If such governance existed for the collection and use of personal data, then the remaining big item is enforcement. You’d hope that the morality and ethics of corporations would be enough to ensure they protected the data entrusted to them with the vigor necessary to fight off the vast majority of hackers and organized crime, but this is the real world. Apparently the “big stick” approach needs to be reinforced.

A few months ago I delved in to how the fines being levied against organizations that had been remiss in doing all they could to protect their customer’s personal data should be bigger and divvied up. Essentially I’d argue that half of the fine should be pumped back in to the breached organization and used for increasing their security posture.

Looking at the fines being imposed upon the larger organizations (that could have easily invested more in protecting their customers data prior to their breaches), the amounts are laughable. No noticeable financial pain occurs, so why should we be surprised if (and when) it happens again. I’ve become a firm believer that the fines businesses incur should be based upon a percentage of valuation. Why should a twenty-billion-dollar business face the same fine for losing 200,000,000 personal records as a ten-million-dollar business does for losing 50,000 personal records? If the fine was something like two-percent of valuation, I can tell you that the leadership of both companies would focus more firmly on the task of keeping yours and mine data much safer than they do today. 

-- Gunter Ollmann

First Published: IOActive Blog - March 26, 2014

Thursday, June 20, 2013

FDA Safety Communication for Medical Devices

 The US Food and Drug Agency (FDA) released an important safety communication targeted at medical device manufacturers, hospitals, medical device user facilities, health care IT and procurements staff, along with biomedical engineers in which they warn of risk of failure due to cyberattack – such as through malware or unauthorized access to configuration settings in medical devices and hospital networks.

Have you ever been to view a much anticipated movie based upon an exciting book you happened to have read when you were younger, only to be sorely disappointed by what the director finally pulled together on the big screen? Well that’s how I feel when I read this newest alert from the FDA. Actually it’s not even called an alert… it’s a “Safety Communication”… it’s analogous to Peter Jackson deciding that his own interpretation of JRR Tolkien’s ‘The Hobbit’ wasn’t really worthy of the title so to forestall criticism he named the movie ‘Some Dwarves and a Hobbit do Stuff’.

This particular alert (and I’m calling it an alert because I can’t lower myself to call it a safety communication any longer) is a long time coming. Almost a decade ago me and my teams at the time raised the red flag over the woeful security of hospital networks, then back in 2005 my then research teams raised new red flags related to the encroachment of unsecured WiFi in to medical equipment, for the last couple of years IOActive’s research team have been raising new red flags over the absence of security within implantable medical devices, and then on June 13th 2013 the FDA releases a much watered down alert where the primary recommendations and actions section simply states “[m]any medical devices contain configurable embedded computer systems that can be vulnerable to cybersecurity breaches”. It’s as if the hobbit has been interpreted as a midget with hairy feet.

Yes I joke a little, but I am very disappointed with the status of this alert covering an important topic.

The vulnerabilities being uncovered on a daily basis within hospital networks, medical equipment and implantable devices by professional security teams and researchers are generally more serious than what outsiders give credit. Much of the public cybersecurity discussion as it relates to the medical field to date has been about people hacking hospital data systems for patient records and, most recently, the threat of targeted slayings of people who happen to have vulnerable implanted insulin pumps and heart defibrillators. While both are certainly possible, they’re what I would associate with fringe events.

I believe that the biggest and most likely threats lie in non-malicious actors – the tinkerers, the cyber-crooks, and the “in the wrong place at the wrong time” events. These medical systems are so brittle that even the slightest knock or tire-kicking can cause them to fail. I’ll give you some examples:

  • Wireless heart and drug monitoring stations within emergency wards that have open WiFi connections; where anyone with an iPhone searching for an Internet connection can make an unauthenticated connection and have their web browser bring up the admin portal of the station.
  • Remote surgeon support and web camera interfaces used for emergency operations brought down by everyday botnet malware because someone happened to surf the web one day and hit the wrong site.
  • Internet auditing and scanning services run internationally and encountering medical devices connected directly to the Internet through routable IP addresses – being used as drop-boxes for file sharing groups (oblivious to the fact that it’s a medical device under their control).
  • Common WiFi and Bluetooth auditing tools (available for android smartphones and tablets) identifying medical devices during simple “war driving” exercises and leaving the discovered devices in a hung state.
  • Medial staff’s iPads without authentication or GeoIP-locking of hospital applications that “go missing” or are borrowed by kids and have applications (and games) installed from vendor markets that conflict with the use of the authorized applications.
  • NFC from smartphone’s and payment systems that can record, playback and interfere with the communications of implanted medical devices.

These are really just the day-to-day noise of an Internet connected life – but one that much of the medical industry is currently ill prepared to defend against. Against an experienced attacker or someone determined to cause harm – well, it’s as one sided as a lone hobbit versus the combined armies of Middle Earth.

I will give the alert some credit though, that did clarify a rather important point that may have been a stumbling block for many device vendors in the past:

“The FDA typically does not need to review or approve medical device software changes made solely to strengthen cybersecurity.”

IOActive’s experience when dealing with a multitude of vulnerable medical device manufacturers had often been disheartening in the past. A handful of manufacturers have made great strides in securing their devices and controlling software recently – and there has been a change in the hearts and minds over the last 6 months (pun intended) as more publicity has been drawn to the topic. The medical clients we’ve been working most closely with over recent months have made huge leaps in making their latest devices more secure, and their next generation of devices will be setting the standard for the industry for years to come.

In the meantime though, there’s a tremendous amount of work to be done. The FDA’s alert is significant. It is a formal recognition of the poor state of security within the industry – providing some preliminary guidance. It’s just not quite a call to arms I’d have liked to see after so many years – but I guess they don’t want to raise too much fear, nor the ire of vendors that could face long and costly FDA re‑evaluations of their technologies. Gandalf would be disappointed.

(BTW I actually liked Peter Jackson’s rendition of The Hobbit).

-- Gunter Ollmann

First Published: IOActive Blog - June 20, 2013

Wednesday, May 29, 2013

Security 101: Machine Learning and Big Data

 The other week I was invited to keynote at the ISSA CISO Forum on Incident Response in Dallas and in the weeks prior to it I was struggling to decide upon what angle I should take. Should I be funny, irreverent, diplomatic, or analytical? Should I plaster slides with the last quarter’s worth of threat statistics, breach metrics, and headline news? Should I quip some anecdote and hope the attending CISO’s would have an epiphany that’ll fundamentally change the way they secure their organizations?

In the end I did none of that… instead I decided to pull apart the latest batch of security buzzwords – “Big Data” and “Machine Learning”.

If you attended RSA USA (or any major security vendor/booth conference) this year you can’t have missed the fact that everything from Antivirus through to USB memory sticks now come with a dab of big data, a sprinkling of machine learning, and a dollop of cloud for good measure. Thankfully I’m a cynic; or else I’d have been thrashing around on the ground in an epileptic fit from all the flashy marketing claims and trademarked nonsense phrases.

I guess I’m lucky to be in the position of having had several years of exposure to some of the greatest minds at Georgia Tech as they drummed in to me on a daily basis the “what and how” of machine learning in the context of solving many of today’s toughest security problems.

So, it was with that in mind that I thought “If I’m a CISO and everything I know about machine learning and big data came from carefully rehearsed vendor sound bites and glossy pamphlets, would I be able to tell the difference between Chanel #5 and cow manure?” The obvious answer would result in some very happy farmers.

What was the net result of this self-inflection and impending deadline? I crafted a short presentation for CISO’s… a 101 course on machine learning and big data… and it included ducks.

If you’re in the upper tiers of your organization and you’ve had sales folks pimping you their latest cloud-infused, big data-munching, machine learning, and world-hunger-solving security solution, please carry on reading as I attempt to explain the basics of the latest and greatest in buzzwords…

First of all – some context! In the world of breach detection and incident response there’s a common idiom: “If it walks like a duck, flies like a duck, and quacks like a duck… it must be a duck.”


Now I could easily spend another 5,000 words explaining why such an idiom doesn’t apply to modern security threats, targeted attacks and advanced persistent threats, but you’ll have to wait for a different blog post. Rather, for this 101 lesson, it’s important to understand the concept of “Feature Selection” – which in the case of this idiom includes: walking, flying and quacking.

If you’ve been tasked with dealing with a duck problem, ideally you’d be an aficionado on the feet, wings and sounds of ducks. You’d be able to apply this knowledge to each bird you have the time to focus your attention on and make a determination: Duck, or Not a Duck. As a security professional, you’d be versed in the various attributes of certain threats – and able to derive a conclusion as to the nature of the security problem.

The problem though is that at scale things break down.


What do you do when there’s too many to analyze, when time is too short, and when you can’t make out all the duck features from afar? This is typical of the big data problem (and your everyday network traffic). Storing the data is the easy part. Retrieving the data is mechanically complicated, but solvable.

Meanwhile, making decisions and taking actions upon the data is typically the most difficult part. With every doubling of data, your problem grows exponentially.


The traditional method of dealing with the situation has been to employ signature matching systems. In essence, we build rules based upon the features we’ve previously identified as significant and capable of bounding the problem (or duck selection). We then compare these rules against the sample animal and receive a binary answer – Duck, or Not a Duck.

Signature systems can be very efficient at classification. Just look at your average Intrusion Prevention System (IPS). A problem though lies in the scope of the features that had been defined.

If those features (or parameters) used for classification are too narrow (or too broad) then evasion is not only probable, but guaranteed. In essence, for a threat (or duck) to be classified, it must have been observed in the past or carefully predicted (although rare).

From an attacker’s perspective, knowledge of those features and triggering parameters makes it a trivial task to evade or to conduct false flag operations. Just think – hunters do this all the time with their floating duck decoys. Even fellow duck hunters have been known to mistakenly take pot-shots at them too.

Switching pace a little, let’s look at the network a little.

The big green blob is all the network traffic for an organization for a week. The red blog right-of-center is traffic associated with an active breach, and the lighter red blob with the dotted lines are just general malicious traffic observed within the network. In this two-dimensional view (if I hadn’t color-coded it previously) you’d have a near impossible task differentiating between them. As it is, the malicious traffic is mixed with both the “safe” and “breach” traffic.

The trick in differentiating between the network traffic types lies in increasing the dimensionality of the problem. What was a two-dimensional blob suddenly becomes much clearer when an appropriate view or perspective to the data is added. In the context of the above diagram, the addition of a z-axis and an extension in to the third-dimension allows the observer (i.e. analyst) to easily differentiate between the traffic types – for example, the axis could represent “country code of destination IP address”. In this context, the appropriate feature selection can greatly simplify the detection problem. Choosing appropriate features is important – nay, it’s critical!

This is where advances in machine learning over the last half-decade have really come to the fore in computer science and more recently in information security.

Without getting in to any of the math behind the various machine learning algorithms or techniques, the key concept you need to understand is “training”. It can mean many things to many a mathematician, but since we’re likely not one of those, what training means in our context is that we already have samples of what we’re going to be looking for, and samples of things we know we’re definitely not interested in. The better we define and populate these training sets, the more precise the machine learning system we’re employing will be in differentiating between them – and potentially classifying other contenders.

So, in this example we’ve taken a bunch of ducks and grouped them together. They become our “+ve class” – which basically means these are the things we’re interested in. But, equally important, is our “-ve class” – our collection of things we know not to be ducks. In many cases our -ve class may be more important than our +ve class because it contains all those false positives and “nearly” ducks – the things that may have caught us out once before.

One function of a good machine learning system is to automatically determine which attributes make the most sense in differentiating between your +ve and -ve classes.

While our poor old hunter (or analyst) was working with three features – walks, flies, and talks – the computer-based system may have reviewed all the attributes that were available and determined which ones are the most useful in differentiating between “ducks” and “not ducks”. In many cases the system will have weighted the various features (or attributes) to indicate which features are more deterministic of the classes.

For example, texture may be a good indicator of “not a duck” – since none of the +ve class were made from plastic or wood. Meanwhile features such as “wing length” may not be such a good criteria and will be weighted in a way to not have an influence on determining whether a duck is a duck or not – or may be dropped by the system entirely.

The number of features reviewed, assessed and weighted by the machine learning system will eventually be determined by the type of data being used and how “rich” it is. For example, in the network security realm we may be feeding the system with collated samples of firewall logs, DNS traffic samples, IP blacklists/whitelists, IPS alerts, etc. It’s important to note though that the “richer” the data set (i.e. the more possible features there could be), the more complex the problem is for the computer to solve and the longer it’ll take to train the system.

Now let’s switch back to that “big data” we originally talked about. In the duck realm we’re talking about all the birds within a national park (say). Meanwhile, in the network security realm, we may be talking about all the traffic observed in real-time across a corporate network and all the alerting instrumentation (e.g. firewalls, IPS, etc.)

I’m going to do some hand-waving here because it can get rather complex and there’s a lot of proprietary tweaks that can be undertaken here… but in one representation we can get our trained system to automatically group and cluster events on our network.

Using our original training data, or some other previously labeled datasets, it’s possible to automatically label the clusters output by a machine learning system.

For example, in the graphic above we see a number of unique clusters (or blobs if you insist). Through automatic labeling we know that the green blobs are types of ducks, the red blobs are various groupings of not ducks, and the gray blobs are clusters of previously unknown or unlabeled clusters – each one mathematically distinct from the other – based upon the features the system chose.

What the system can also do is assign a probability that the unknown clusters are associated with our +ve or -ve training sets. For example, in this particular graphical representation the proximity of the unlabeled clusters to labeled (and classified) clusters allows the system to assign a probability of whether the cluster is a duck or not a duck – even though the system had never seen these things before (i.e. “birds” the system hasn’t encountered before).

The next (and near final) stage is to manually label these new clusters. For example, we ask an ornithologist to look at each cluster of “ducks” and “not ducks” in turn and to label them… “rubber duckies”, “robot duckies”, and “Madagascar mallard ducks”.

Then, to improve our machine learning system further, we add these newly labeled clusters to our +ve and -ve training sets… and the system continues to learn and become more precise over time.

In addition, since we’ve now labeled these clusters, in the future we’re able to automatically flag new additions to these clusters and correctly label the duck (or threat).

And, if we’re a really smart CISO, we can use this clustering system (and labeled clusters) to automatically alert us to new threats or to initiate automatic network security actions – e.g. enable blocking of a new malicious URL, stop blocking a new cloud service delivering official updates to applications, etc.

The application of machine learning techniques to the toughest security problems facing business today has come along in leaps and bounds recently. However as with any buzz word that falls in to the hands of marketers and gets manipulated until it squeaks and glitters, or oozes onto every product in this year’s price list, senior technical staff need to take added care not to be bamboozled by well-practiced but meaningless word salad.

A little understanding of the concepts behind big data and machine learning can not only cut through the latest batch of sales promises, but can also form the basis of constructing a new generation of meaningful breach detection and incident response solutions.

-- Gunter Ollmann

Original Publication: IOActive Blog - May 29, 2013