Results

Will AI surpass the world’s best hackers? What enables humans to discover, exploit, and patch vulnerabilities—and where does AI already outperform us, and where does it still fall short? To explore this divide and glimpse the future, I’ll compare how we train top human talent with how we train frontier AI models—and whether the goals are even the same. Yet autonomy and intelligence are only part of the puzzle: new technologies matter only if they can be adopted, and the real barriers to adoption are rarely the obvious ones. Drawing on two decades of experience—from research and winning competition systems to commercializing Mayhem technology —I’ll share the hard-earned lessons, surprises, and setbacks that point toward the future of automatically checking and protecting the world’s systems from attackers.

Last year, we discovered the first known security vulnerability in Google’s proprietary DSP coprocessor used in the Pixel 8. Google patched the issue earlier this year. This DSP is undocumented and had never been publicly analyzed before, presenting a significant challenge to reverse engineering and exploitation.
Although our initial attempts to emulate the processor were unsuccessful, we persisted by using dynamic instrumentation techniques and other tricks to gain insight into its behavior. Through this effort, we uncovered a critical bug that allowed us to achieve full kernel code execution and bypass all mitigations on the Pixel 8 including MTE.To build a complete exploit chain, we also identified and chained two additional vulnerabilities, enabling exploitation from a less privileged context. Our research demonstrates a deep exploration into a previously opaque component, and reveals new attack surfaces in modern Android hardware.

Over the past year, we identified and reported 13 security vulnerabilities (CVEs) in Samsung’s Exynos SoC, primarily affecting the NPU and GPU subsystems. These bugs—many of which were exploitable—have since been patched by Samsung as of June 2025. In this talk, we focus on the Exynos NPU: we’ll walk through recurring bug patterns, analyze the exposed attack surface, and explore why these coprocessors remain an attractive target. We will then dive into the exploitation side—demonstrating how some of these vulnerabilities can be leveraged to gain arbitrary read/write (AARW) on devices like the Galaxy A35, A55, and S24+. Beyond that, we’ll show how to chain primitives to bypass the hypervisor and escalate to full kernel code execution The session also reflects on our experience collaborating with Samsung’s product security team, and offers an honest perspective on working through the ISVP, including incentives, technical feedback, and timelines.

Existing LLM-based penetration testing research has achieved some success by designing multi-agent systems that combine autonomous thinking and behavior. However, most of this research has been limited to testing in virtual lab environments, and quantitative evaluation of their application to real-world environments and the automation of all phases from reconnaissance to report generation is insufficient.
This research aims to achieve practical-level penetration testing automation, including reconnaissance, vulnerability analysis, exploitation, and report generation. Based on findings gained from last year’s collaborative research with the Takagi Laboratory at Meiji University (arXiv:2502.15506v1), we developed a multi-agent system using modern Agent AI. For the evaluation system, we collaborated with penetration testers with extensive field experience to create a practical environment.
This presentation will present the latest trends in AI-based penetration testing automation and the specific benefits of its implementation. First, we will outline the research trends since 2023, when LLM was introduced, and in the main part, we will report on a performance evaluation of the constructed system compared with conventional tools and manual diagnosis. In addition to HackTheBox, the evaluation will use an environment simulating a real-world situation to quantitatively compare the degree of reproducibility, coverage, and efficiency. The presentation will also touch on the potential for automation using local LLM. Through this presentation, participants will gain a deeper understanding of the benefits and challenges of introducing AI agents and gain concrete guidelines for applying them to their own company’s security assessments.

Amazon’s Kindle is the most popular e-reader on the market, with an extensive ecosystem of e-books. From a security perspective, Kindle devices especially stand out because they are often linked to an Amazon account.
Their complex software stack supports numerous e-book file formats (AZW, MOBI, PDF…), as well as many underlying media formats that increase the attack surface. As such, downloading an e-book from the store may allow an attacker to gain root access to the device, take control of the Amazon account, and steal credit card information.
In this talk, we will dive into the internals of Kindle devices and discuss a $20,000 bug in the parsing of Audible audiobooks which allowed us to take full control of the e-reader. We will also share general insights on fuzzing file formats based on the MPEG-4 standard (ISOBMFF).

Kernel Streaming emerged as a new attack surface in the Windows kernel last year, leading to multiple in-the-wild exploits. Over the past year we uncovered “proxying to kernel”, a logical bug class that bypasses many privilege checks, making exploitation straightforward. However, this is only the tip of the iceberg for Kernel Streaming.
This time, we focus on one of the most common inputs to Kernel Streaming—frames from a webcam. To boost performance, Kernel Streaming introduced an MDL cache mechanism, but it also opened new vulnerabilities. In this talk, we’ll reveal a new array of bug classes, with 10+ vulnerabilities. We’ll explain the design flaws behind them, why they may look unexploitable at first, and how we turn some into arbitrary physical-memory writes.
By witnessing the power of these bugs, attendees will be able to discover and defend against more local privilege-escalation flaws in Windows.

This research explores macOS IPC security, specifically focusing on Mach message handlers in system daemons. These handlers, which expose privileged functionality, present a significant attack surface for sandbox escapes and local privilege escalation.
I’ll demonstrate how I used structured fuzzing and a technique called API call chaining to uncover vulnerabilities in the coreaudiod system daemon on MacOS. My custom fuzzing harness, dynamic instrumentation, and a blend of static and runtime analysis led to several security flaws, including two major memory corruption bugs. I’ll detail the full exploit chain I developed to leverage one of these for a sandbox escape on modern macOS.
I’ll also discuss challenges faced, such as initializing CoreAudio, mocking components, and building targeted grammars for fuzzing. Finally, I’ll share the open-source fuzzing harness and tools developed during this research, aiming to enhance macOS IPC fuzzing accessibility for the security community.

Propose a new generalized precise Fault-Injection methodology, that can work against any micro-controllers.
The new methodology makes a precise Fault-Injection attack significantly easier, by eliminating the needs of implementing a custom application-level communication driver before an attack. Usually, such a driver is required to capture both a digital trigger and a success signal, and is also used to issue further commands after a successful glitch. Such a driver is vendor/chip dependent, and usually takes immense research/reverse/debug effort to implement.
The new method did everything by replaying legitimate communications, which are easily obtainable by capturing the communication between a new chip and an official debugger.
PoC: dumped firmware from RL78, a micro-controller used in PS4 and cars, used undocumented commands, bypassed On-Chip debug security ID and more. The attack is stable.
Implemented with a “Pico” micro-controller. All code is on the author’s github.

This talk covers the PerfektBlue attack - a set of critical memory corruption and logic vulnerabilities found by Mikhail Evdokimov in BlueSDK Bluetooth stack that can be chained together to obtain over-the-air Remote Code Execution (RCE) on millions of vehicles manufactured by different vendors. You will be guided through the entire vulnerability research process, from the initial discovery to building a sophisticated exploitation chain to get RCE on multiple targets. The session includes an overview of Volkswagen, Mercedes-Benz, and Skoda IVI systems, followed by an introduction to the Bluetooth architecture. Further, we delve into the discovery and exploitation phases, describing how a UAF vulnerability can be turned into the Arbitrary Address Write (AAW) primitive, along with the memory leak via the logic bugs’ chain. After this, we put it all together to obtain the execution of arbitrary functions leading to RCE on IVI systems.

The finals of the AI Cyber Challenge (AIxCC) were held this summer at DEF CON 33 in Las Vegas, USA.
This competition is a two-year national project led by DARPA (Defense Advanced Research Projects Agency) and **ARPA-H (Advanced Research Projects Agency for Health), aiming for the convergence of AI and cybersecurity. Its goal is the automated defense of software supporting critical infrastructure. Participating teams developed AI-powered systems for vulnerability detection and remediation and competed to demonstrate their performance.
This panel discussion will feature David Brumley (Keynote Speaker), who contributed to the organization of the AIxCC, as well as members of the teams that placed highly in the finals. They will exchange views on the forefront and future of AI and cybersecurity.

During war or disasters, temporary evacuation shelters and volunteer field hospitals become the softest cyber targets. Yet they still need Wi-Fi for MyNumber identity checks, EMR exchange and supply tracking, while lacking SOC staff, stable power or bandwidth. We present Azazel System, an open-source “Cyber Scapegoat Gateway” on a single Raspberry Pi 5 that boots a full SOC/NOC—Suricata IDS/IPS, OpenCanary decoys, Vector log pipeline, Mattermost alerts—within 15 minutes. We share architecture, field tests (30 s detection, 12 % compromise rate, 13 W), and the legal model that lets volunteers run it under Japan’s new Active Cyber Defense rules. Takeaways: ① build & image the device, ② tune latency-injection to slow attackers, ③ integrate with 00000JAPAN or LEO-satellite links. Live demo and open-source image provided.

The initial phase of binary analysis-understanding “what a program actually does”-is fundamentally broken. Static analysis shows code that might run, while debuggers offer a slow, step-by-step view of a single execution path, failing to capture the dynamic, multi-threaded reality of modern software. This creates a critical bottleneck for triage and incident response.
We present BIN2TL, a lightweight, high-level execution tracer, not another debugger. Using Intel Pin, it captures key events (function calls, thread activity) from concrete execution and converts them into a standard Perfetto timeline. The result is a complete, interactive, high-level map of the program’s behavior over time.
This approach provides what other tools cannot: a rapid, holistic overview. We will demonstrate how BIN2TL makes complex analysis intuitive. See how ransomware encryption threads operate in parallel, or instantly identify the code regions used by a specific feature.

This talk presents an open-source lab designed for those starting out with internal network testing. Rather than a hands-on workshop, the session focuses on how this environment can be used as a self-paced tool to practice port forwarding, bypassing network segmentation, and lateral movement using tools like Chisel and Ligolo-NG. By the end of the session, attendees will know how to deploy the lab, explore different pivoting scenarios, and contribute to the project if they choose. It’s a practical resource for anyone looking to sharpen their offensive security skills or set up realistic testing environments with minimal hassle.

In embedded systems, various CPU architectures are used. On the other hand, one of the software vulnerability countermeasures is CPU-based security measures. This can involve adding dedicated security instructions or designing and developing new CPU architectures with security in mind, but developing compilers that support these becomes a barrier. Furthermore, compiler security mechanisms, represented by stack protectors, are effective because they can be applied without modifying the source code, but verifying and introducing new security mechanisms requires compiler modification. Moreover, when performing vulnerability analysis on a new CPU, understanding the CPU’s operation is necessary, and for that, the existence of a compiler that can generate executable code for that CPU is crucial.
For these reasons, a compiler that can quickly adapt to new CPU architectures is required. While some conventional open-source compilers, such as GCC, support multiple CPU architectures, many of them are highly optimized, making it extremely difficult and impractical to adapt them to new CPU architectures.
Therefore, we propose the Reduced Assembly Set Compiler (RASC), which facilitates adaptation to new CPU architectures by limiting the patterns of assembly to be generated, and have developed the C compiler NLCC as an implementation of RASC.
NLCC can support new CPU architectures by simply registering 64 types of assembly patterns, and furthermore, 24 types of patterns can be omitted by built-in functions. As a result of adapting to 12 types of CPU architectures, including x86 and ARM, it was possible to adapt in an average of about 8 hours, making it usable as a “1-day compiler” that can support new CPU architectures in one day. We will examine and discuss the possibilities of RASC and the security functions in NLCC.

While the importance of Red Team exercises has been increasing in recent years, the process of developing attack scenarios, preparing tools, and building execution environments requires a significant amount of time and effort. This poses a major challenge that limits the frequency and quality of exercises.
To address these challenges, we have developed “GHARF,” an innovative framework that applies the mechanisms of Continuous Integration / Continuous Delivery (CI/CD) to Red Team operations, enabling efficient exercise execution.
This tool automates various phases, from the development of simulated attacks to their preparation and execution, by applying the CI/CD build and delivery mechanisms to Red Team operations. This significantly improves the efficiency of Red Team operations and enables rapid operational cycles. We call this concept “Continuous Attack Integration / Continuous Attack Delivery, Deployment (CAI/CAD).”
Tools offering similar functionalities include MITRE CALDERA and Atomic Red Team, which are categorized as BAS (Breach and Attack Simulation) tools. These tools aim to reduce Red Team workload and enable self-assessment by Blue Teams through the automation of simulated attacks. In contrast, GHARF is characterized by its focus on improving the efficiency of Red Teams themselves when conducting operations. Our approach aims to optimize the process for Red Teams to pursue more advanced and practical attack scenarios.
GHARF, at present, is limited to conveying the concept from an ethical perspective. However, we believe that by presenting concrete examples of CAI/CAD through this presentation, this concept will spread and contribute to the further development of the Red Team field.
Details of the tool are available in the README of the GitHub repository below. Please refer to it.
https://github.com/nttcom/gharf

[Recorded Session]
We’ve been tracking a sophisticated state-sponsored campaign since Feb 2025, targeting a South American foreign ministry before spreading to Southeast Asia. Our research uncovered novel malware families, notably a modular, cross-platform backdoor called FINALDRAFT, which uses Microsoft Graph API for C2. Despite its sophistication, the operators made key OPSEC errors, exposing infrastructure and pre-release malware.
We’ll present technical insights into the malware’s evolution, custom protocols, and FINALDRAFT’s modules for lateral movement, script execution, and enumeration. The campaign continues as we observe recent activity from April to June 2025 leveraging open-source tooling, obfuscation, and the use of both offensive security tools and highly obfuscated malware.
Tailored for researchers, SOCs, and AV vendors, this talk provides actionable intelligence and covers this group TTPs. A custom tool will be released to interact with the malware and aid detection development.

In an era where synthetic media and deepfakes are becoming tools of choice for adversaries, this session delivers a deep dive into the entire lifecycle of a synthetic media attack—from initial OSINT gathering to monetization through fraud and extortion. Drawing from real-world incidents, cutting-edge research, and red-team simulations, we’ll dissect how deepfake-based attacks are operationalized, bypass controls, and reshape the threat landscape across sectors.
Finally, the session presents a comprehensive defense framework—covering AI-driven detection techniques, content authenticity infrastructure (C2PA), security engineering controls, and organizational playbooks for executive impersonation response. By the end of this session, security professionals, risk leaders, and technical architects will be equipped with actionable strategies to detect, disrupt, and defend against synthetic media threats in the real world.

Cloud platforms rely on deep trust between internal components—but what if that trust is broken? This talk presents the discovery and exploitation of CVE-2025-29972, a critical 9.9 CVSS flaw in Azure’s Storage Resource Provider. We reveal a multi-stage attack chain starting from a classic SSRF, used to hijack the identity metadata fetching process and leak Azure Active Directory (AAD) tokens for arbitrary tenants.
Our research introduces “Spray&Pray4Bind,” a novel DNS rebinding technique built to bypass modern caching defenses that render traditional rebinding ineffective. We walk through the full exploit: from SSRF to token abuse, lateral movement, and ultimately regenerating SFTP passwords—compromising tenant storage. Based on internal offensive research at Microsoft, this talk shows how broken trust in cloud services can lead to full compromise—and provides defensive insights for securing complex cloud environments.

Memory Corruption is not commonly associated with Cloud Security. While taken seriously, it is a theoretical risk that is rarely reported to be exploited successfully. We believe that there are multiple reasons: Cloud services are typically written in Memory Safe languages and run behind Load Balancers that introduce variability that defeats common Exploit techniques. Finally, attackers are missing crucial information about the binary they are targeting, such as offsets for ROP chains etc.
In this talk, we explain how attacking Cloud services differs from conventional Memory Corruption targets and challenges that attackers need to overcome. We then go in-depth into an end-to-end vulnerability chain that resulted in Remote-Code-Execution on Google Cloud’s Artifact Analysis backend.
Our goal for this talk is to demonstrate that exploiting Memory Corruption vulnerabilities is feasible, even when using Memory Safe languages and without attacker knowledge of the backend binary.

Over the past two years, a new class of attacks known as Promptware has emerged, exploiting LLMs at inference time via crafted prompts. Though often dismissed as impractical or exotic, this talk will shatter that misconception forever. We introduce Targeted Promptware Attacks, where an attacker invites a victim to a Google Calendar meeting containing an indirect prompt injection. This hijacks Gemini’s integrated agents, on web, mobile, and Google Assistant-which operate with OS-level Android permissions. We demonstrate 15 real-world exploits, including spamming, phishing, data exfiltration, calendar deletion, device control (e.g., boiler, lights, windows), video streaming a victim via Zoom, and geolocating the victim. These attacks show Promptware’s ability to move laterally across agents and devices, leading to physical-world consequences. Using our threat assessment framework, we find that 73% of identified risks are high-critical, calling for immediate mitigations.

North Korea deploys sophisticated cyber operations to generate foreign currency through cryptocurrency theft and covert IT worker placements. These funds directly support the Kim regime’s power consolidation and nuclear weapons development.
Our investigation provides unprecedented visibility into these operations’ human elements and organizational structures. Unlike previous research that focused on technical indicators or theoretical attribution, we reveal the operational workflow through advanced OSINT techniques—from sophisticated identity forgery and cover story development to command hierarchies and field operations.
We present actionable intelligence, including social engineering patterns, fake ID creation methods, and detailed playbooks for cultivating cover accounts. This intelligence equips security professionals with practical countermeasures against these sophisticated threat actors and offers rare insights into the actual mechanics of North Korean cyber operations.

The presentation will cover North Korean remote IT workers and the broad operational methods they utilize. It will start with their likely methods of faking, stealing, or purchasing credentials or identities, and some ways they fake their identity and background in an interview. Next, it will cover their malicious actions once they gain employment, including deploying ransomware or malware, or stealing information from the employer to sell on the dark web. There will be a section on their preferred targets and how to spot a “laptop farm” that North Koreans set up through their proxies overseas. The presentation will also include details on the software they use to obfuscate their identity and real location, as well as an overview of insights gleaned from the data logs of a suspected North Korean IT worker’s computer. It will end with a discussion on some mitigations to prevent companies from unintentionally hiring North Korean IT workers.

Japan’s Cyber Counter-capacity Enhancement Law (CCEL) authorises active cyber defence while remaining within constitutional limits. It enables preventive domestic actions—such as malware neutralisation and IP tracing—under civilian oversight, without amounting to offensive cyber warfare. The law’s ambiguity on attribution and pre-emption raises critical legal and geopolitical questions. This talk examines the CCEL’s innovative approach to cyber sovereignty, international cooperation, and the legality of pre-emptive measures. It offers a comparative analysis with Western frameworks and explores its relevance for global cyber operations and multilateral governance.

In Japan, the Cyber Response Capability Enhancement Act and its implementation Act were enacted in May 2025. A notable aspect of the law is the “Access & Neutralization Measures,” which stipulates that the Cyber Harm Prevention Enforcement Officer, under the Police Duties Execution Act, can independently take “necessary measures deemed ordinarily required” to delete harmful electromagnetic records or take other harm prevention actions (as per Article 6-2 of the law). However, there is ambiguity regarding the specific actions permitted under this law, as well as the detailed procedures and associated challenges. To address these uncertainties, legal experts in Japan and policy specialists from the U.S discuss in following issues;
1.Overview of the Cyber Response Capability Enhancement Act and Implementation Act
2/Offensive Cyber Operations in the U.S.
3.Comparison
4.Suggestions for Japan

In Japan, legislation to introduce “active cyber defense” was enacted in May of this year, and the National Cyber ​​Security Office was established in July. Currently, the government is working to establish a system for implementing active cyber defense and to formulate a cybersecurity strategy. This presentation will introduce the Japanese government’s policy trends, including human resource development, and how to promote measures to strengthen cybersecurity and ensure the safety of cyberspace, which will serve as the foundation for supporting society in the future.

Reverse engineering rootkits is increasingly challenged by advanced obfuscation and packing, hindering dynamic debugging of Windows drivers. While Unicorn-based frameworks like Speakeasy and Qiling exist, they are still insufficient in anti-simulation techniques.
This research proposes a Unicorn-based semi-simulation framework that executes drivers in a hybrid real-simulated environment via partial pass-through, extracting real environment components and supporting parallel execution and structure exception handling to bypass anti-simulation and anti-debugging protections. Running isolated in Ring 3, it can precisely monitor objects and registers, revealing rootkits’ logic and its self-protect mechanisms.
I will explore modern anti-debugging techniques, Unicorn applications, and a case study of a high-market-share anti-cheat engine’s kernel driver protections. After this session, attendees will gain a better understanding of internal driver protection and rootkit analysis.

Bootkits and rootkits are stealthy malware targeting the lowest system layers to bypass security defenses. Bootkits alter firmware or bootloaders to seize control early in boot, while rootkits hide in the OS kernel to conceal activity. Yet research is limited by complexity and lack of real samples.
In this talk, we present BOOTKITTY, a hybrid bootkit‑rootkit framework for Windows, Linux, and Android. By analyzing each OS’s secure boot mechanisms, we show how to generalize bootkit attacks to defeat platform‑specific protections. We demonstrate exploiting UEFI drivers and bootloader vulnerabilities to break the trust chain and bypass mitigations.
Our findings reveal systemic weaknesses in boot security and highlight the feasibility of cross‑platform bootkit attacks. This underscores the urgent need for unified, resilient secure boot designs.

CICDGuard is a graph based CICD ecosystem visualizer and security analyzer -
1. Represents entire CICD ecosystem in graph form, providing intuitive visibility and solving the awareness problem
2. Identifies common security flaws across supported technologies and provides industry best practices for identified flaws adhering to OWASP CICD Top10
3. Identifies the relationship between different technologies and demonstrates how vulnerability in one component can affect one or more other technologies
4. Technologies supported - GitHub, GitHub Action, Jenkins, JFrog, Spinnaker, Drone
CICD platforms are an integral part of the overall software supply chain and it processes a lot of sensitive data, compromise of which can affect the entire organization. One of the challenges with security OF CICD, like most areas of security, is the lack of visibility of what actually makes a CICD ecosystem. Security starts with being aware of what needs to be secure.

Windows Event Logs play a critical role in DFIR. However, default audit settings often lead to detection blind spots due to limited log sizes, insufficient policies, and short retention.
We present two open-source tools that help defenders assess and improve Windows Event Log audit settings for stronger threat detection and forensic readiness.
WELA is a PowerShell-based tool that audits current settings against industry best practices and real-world Sigma rule coverage. By leveraging over 4,000 Sigma rules, it reveals what threats can or cannot be detected and supports multiple audit guide.
EventLog-Baseline-Guide is a Streamlit-based web application for visualizing differences in audit policies across baseline guides. It uses intuitive color coding and maps audit settings to Sigma rule coverage by log source and event type.
Together, these tools empower security teams to close visibility gaps in Windows logging and make informed, data-driven improvements to their DFIR posture.

This presentation introduces “TOAMI,” a browser extension tool designed to assist phishing hunters.
TOAMI is an open-source tool developed for phishing hunters to detect phishing sites in real-time. Unlike traditional phishing site detection, which primarily focuses on identifying corporate brand misuse, TOAMI analyzes phishing kits using Indicators of Kit (IoK) to identify malicious sites.
As a browser extension, TOAMI enables anyone to easily conduct site investigations with minimal setup.
When TOAMI is activated, investigation logs, site screenshots, and other relevant data are automatically collected and saved locally. Phishing sites often employ techniques like cloaking to prevent re-access to the same site. TOAMI’s functionality allows it to capture crucial evidence, such as screenshots, even when cloaking techniques are used, thereby avoiding issues like inaccessible sites or missing data.

In digital forensics, constructing filesystem timelines is crucial. However, conventional methods rely on MACB timestamps at disk image acquisition, which are limited to recent states and vulnerable to timestomping—and critical activity may be missed. Ext4 and XFS use journaling to protect against system crashes. These journals record low-level operations that, if decoded, provide chronological traces of past file activity without relying on standard metadata that are easier to modify. Despite open specifications, practical forensic tools for analyzing ext4 and XFS journals have been lacking. In this talk, I will introduce FJTA (Forensic Journal Timeline Analyzer), a new open-source tool designed to extract file activity from ext4 and XFS journals. I will explain journal structures, demonstrate how file operations can be reconstructed, and showcase examples of how this technique can supplement conventional forensic evidence—especially when timestamps are untrustworthy.

In recent years, malware analysts and security researchers have actively utilized signature-based formats such as YARA and SIGMA to aid in threat detection and hunting. However, existing endpoint security tools often lack the ability to directly leverage YARA and SIGMA, as they rely on their own proprietary detection engines.
Furthermore, kernel-level drivers for endpoint protection carry the risk of compromising OS stability. To address this, we developed YAMAGoya, a new Threat Hunting tool that operates solely in userland without requiring kernel drivers.
YAMAGoya supports SIGMA signatures and monitors a wide range of events including files, processes, registries, and network communications. It also implements a memory scanning function using YARA, enabling more precise malware detection. Designed for both GUI and command-line operation, it caters to a wide range of use cases, from security operations teams to individual researchers.
This session will detail how YAMAGoya achieves threat detection on Windows using YARA and SIGMA signatures. Through a demonstration simulating actual attack scenarios, we will specifically show how YAMAGoya visualizes, detects, and contains threats. We believe that this tool, which allows seamless utilization of the vast number of signatures published by the global security research community, will significantly expand the possibilities for endpoint threat countermeasures.