Detection at Scale

Matthew Valites, Director of Threat Detection & Operational Strategy at SAP, discusses the best threat detection approaches, using detection as code, and the role of GenAI in the future security landscape. He also shares actionable lessons from his book, 'Crafting the Infosec Playbook'.

Meta's Justin Anderson discusses how they built a detection platform treating it like software code, gauging risk using TTPs, and taking a shift-left approach. They emphasize the need for strong engineering and investigation skills, AI limitations in detection, and advice for building a security program.

On this week's episode of the Detection at Scale podcast, Jack talks with Charles Anderson, Director, Global SOC at Sony. They discuss better approaches to risk-based alerting that leverage metadata, how they fine tune detections across a global organization, and what factors to use when determining thresholds. They also talk about how to use Time to Detect to improve your strategies, how LLMs can help with baseline detection, and why it's key to not lose sight of risk in pursuit of threat. Topics discussed: A better way to approach risk-based alerting by leveraging metadata to connect the dots. Which factors to consider when determining your thresholds for alerting. How Sony is using machine learning and why applying a single model to the entire organization doesn't work. Why organizations are targets of opportunity and accidental exposure more than they are of planned attack. The process Sony's SOC uses to fine tune their detections and how it has to be different across the globe. How to use Time to Detect to tell the story of what you're covering and what you're missing. Advice to other security professionals that includes not losing sight of risk in pursuit of threat.

In this discussion, Jason Craig, the Director of Threat Detection & Response at Remitly, dives into the TTPs of threat actors like Lapsus$. He advocates for hardware-backed authentication over SMS MFA for stronger identity management. Craig emphasizes the importance of a solid asset inventory and understanding organizational threats before crafting effective threat models. He also offers valuable insights on risk-based approaches to protecting sensitive data and the necessity for behavioral profiling to filter out irrelevant noise in security.

On this week's episode of the Detection at Scale podcast, Jack talks with Drew Gatchell, Director, Detection Engineering at AppOmni. They discuss how to overcome the challenges to detection on SaaS platforms and how they're building strategies upon alerting and detection frameworks. They also talk about how generative AI can help with normalizing inputs, the benefits of data lakes for D&R, and why it's key to have a measurable plan for detection. Topics discussed: How AppOmni is tackling the challenges of detection in SaaS platforms and auto-logs, especially when it comes to varied latency. What frameworks Drew is working with and how he's building upon them for better detection. How signal creation starts with a hypothesis that can be turned into a plan, and why it's important to include signal redundancy. What techniques AppOmni takes to address security in real time. How they're using AI to normalize their inputs and create additional content on top of the detection rules. The benefits of data lakes and how they're a tremendous asset to D&R. Advice for security leaders on having a measurable plan for detection, why detection should be layered, and the need to continuously validate your capabilities.

On this week's episode of the Detection at Scale podcast, Jack talks with Emanueal Mulatu, Senior Engineering Manager - Detection & Response at Block. Together, they discuss what success means in security, the most rewarding things about security, and how to address and prevent one of the biggest challenges today: burnout. They also talk about ways to increase productivity through automation, the potential for AI and large language models, and why creating a great workplace starts with a healthy work-life balance. Topics discussed: The most rewarding things about security — like the relationships and trust you build — and the biggest challenges facing security today. The value of building relationships across departments as well as with your customers. How to recognize the root causes of burnout and address it through meaningful initiatives like fitness or reading challenges. Why having a culture of writing can help with problem solving and collaboration. Why automation is the biggest initiative that's increasing productivity and morale, and the opportunities that AI and LLMs will bring. Advice for security leaders on how to build better workplaces focused on psychological safety and continuous learning. How to define security success, especially through the eyes of the C suite.

On this week's episode of the Detection at Scale podcast, Jack talks with Dr. Anton Chuvakin, Senior Security Staff at the Office of the CISO at Google Cloud. They dig deeper into the conversation taking place online around decoupled SIEMs, which both Jack and Anton wrote about. They discuss what a decoupled SIEM is, the evolution of data platforms and security capabilities, if decoupled SIEMs will work broadly with current customer demands, and if having backend data lakes is the best solution for fast, real-time querying. Topics discussed: What is a decoupled SIEM, and why the broader discussion around whether security data lakes will replace SIEMs prompted Anton's Medium post. How this conversation is being driven by the fact that we’re coming to the "end of the runway" on previous storage choices. The arguments around why decoupling may not work broadly, simply because customers want integrated SIEMs. The evolution of data storage platforms and how successful past attempts at integrating security capabilities were. Why there's not a straightforward solution to storage — and why it's a challenge that's taking years to solve. Why having a data lake on the backend is the best solution to fast querying and real-time detection. A discussion around OCSF and the benefits of log normalization.  Resources Mention:  “Decoupled SIEM: Brilliant or Stupid?” by Anton Chuvakin “The Transition from Monolithic SIEMs to Data Lakes for Security Monitoring” by Jack Naglieri

On this week's episode of the Detection at Scale podcast, Jack talks with Dhruv Majumdar, Director, Cyber Risk & Advisory at Deloitte. They discuss common challenges when transitioning from a traditional SOC to a detection and response program, what questions to ask when building a threat modeling strategy, and the benefits data lakes can unlock for D&R. They also talk about how LLMs are helping detect exfiltration and –the need for security controls, policies, and good partnerships. Topics discussed: The common challenges that organizations face today when evolving their detection and response programs, including moving away from SOC and managing big data. An overview of the maturity model and what organizations can follow to evolve their processes. Two critical questions to ask that will guide your threat modeling strategy. What big data "unlocks" for detection and response today, and what trade-offs there are in usability when moving to a data lake-backed architecture. How LLMs can surface patterns in data that simplify detecting exfiltrations and how it can help with automation to prevent burnout. Advice to security practitioners when transitioning to new strategies, including why you need "controls, controls, controls," and why you should take the simplest route to overcome a challenge.   

On this week's episode of the Detection at Scale podcast, Jack talks with Anton Chuvakin, Security Advisor at the Office of the CISO at Google Cloud, and Timothy Peacock, Senior Product Manager at Google. Together, they discuss some of the needs and trends in cybersecurity today, including how to know what level of D&R your organization needs, the use cases for AI today, and how LLMs and SIEMs will handle data at scale. They also talk about the need for more creative solutions to misconfiguration management, three things security practitioners can do to improve cloud security, and why cybersecurity is the "most intellectually stimulating profession on the planet." Topics discussed: What attracted Anton and Timothy to cybersecurity, what makes them stay, and why the intersection of humans and technology make it the “most intellectually stimulating profession on the planet.” How organizations can evaluate the level of security they need, why it's crucial to know whether you need to go from zero to one, or five, or a hundred, and how organizations with no detection and response strategies can get started. What use cases there will be for AI in cybersecurity, and while it may be good at summarizing, explaining complexity, and classifying, it may not be ready to create usable code. Why security practitioners need to think more about whether SIEMs can support planetary scale, and whether decentralization is the solution.  The role LLMs will play in helping to manage large data sets, and how it may change the way organizations use MDRs. Why the industry needs new, creative ways to solve the ongoing problem of cloud misconfigurations in order to break vicious cycles through shared faith.  Three pieces of advice to improve cloud security, including knowing your security needs, practicing, and making friends so you know you're note alone.

In this episode, Jack speaks with David Seidman, Head of Detection and Response at Robinhood. David has worked for large tech companies like Google, Microsoft, and Salesforce in a variety of D&R roles.  During this episode, David shares his tactical advice on how his team is building the pipes and engines of security at Robinhood, his top tools to improve fidelity of detections, and what he’s learned in his career that’s made him a better practitioner and leader.  Topics discussed:  The ‘unusual strategies’ and hypothesis on the kill chain model David has not shared before publicly  His top five tools to use to improve the fidelity of your detections  How David has seen composite detection be effective in practice and why it is most effective when it’s analyst driven  His experience working on Google Cloud's Event Threat Detection  What a mature IR process look like today and how to train staff that’s run IR in the past A big challenge and growth area in the industry that doesn’t get enough attention  The new frontier of what the detection and response stack will look like in the future David’s keys to an effective IR program, such as regular exercises, communications plan, having access and permissions to data, strong controls, and more.   The three actionable takeaways David learned from his roles at Google, Microsoft, Salesforce, and now Robinhood that make him a better practitioner and leader today