appsec.fyi

Talks — A Practical Guide

A curated AppSec resource library covering XSS, SQLi, SSRF, IDOR, RCE, XXE, OSINT, and more.

Talks: A Practical Guide

Curated and synthesized by . Last updated 2026-11-01. Synthesized from 91 of 91 curated resources. Browse all 91 Talks resources →

Problem Framing: The Evolving Threat Landscape

The application security landscape is in constant flux, presenting new challenges and opportunities for practitioners. Recent conferences and research highlight a rapid evolution driven by factors such as the increasing sophistication of attackers, the pervasive integration of Artificial Intelligence (AI) into development and security tooling, and the ever-growing complexity of modern application architectures. Understanding these shifts is critical for developing effective defensive strategies.

A significant trend is the elevation of seemingly isolated vulnerabilities into critical attack chains. A real-world penetration test of a global platform with over 30 million users demonstrated how exposed secrets in JavaScript, session abuse, cross-application trust violations, and misconfigured CORS and WAFs, when combined, could lead to full system compromise, including the exposure of production credentials and cryptographic keys [1]. This highlights the importance of viewing findings not in isolation but as potential components of a larger attack narrative.

The emergence of AI, particularly Large Language Models (LLMs), presents a dual-edged sword. On one hand, LLMs are demonstrating remarkable capabilities in discovering vulnerabilities, automating repetitive tasks, and even generating exploit code [2][3][4][5]. This has led to discussions about the impact of AI on CTF competitions, where previously challenging puzzles are now being solved rapidly through automated LLM workflows [3]. On the other hand, AI also introduces new attack vectors. AI coding agents can be hijacked through poisoned input, leading to code execution and credential theft [6]. Chatbots, when integrated into customer service, can be weaponized via email to send phishing messages or exfiltrate sensitive data [7]. Furthermore, the rise of "Shadow AI" – unsanctioned AI tools in use within organizations – poses a significant and growing risk that security teams are struggling to track [8].

The complexity of modern software development also demands attention. Applications are no longer monolithic but are composed of microservices, containers, multi-cloud environments, and a mix of internally developed, AI-generated, and third-party code [9]. This intricate web of dependencies and deployments increases the attack surface and complicates security efforts. Securing these complex environments requires a shift towards proactive, developer-centric security practices and the intelligent use of tools that integrate seamlessly into the Software Development Lifecycle (SDLC) [9].

Core Mechanics: Attack Vectors and Exploitation Patterns

Understanding the fundamental mechanics of how vulnerabilities are discovered and exploited remains paramount, even with the advent of new technologies. Core attack patterns, while sometimes applied in novel ways, continue to be the bedrock of many breaches.

Injection Attacks: An Enduring Threat

Injection attacks, a class of vulnerabilities that has persisted for decades, continue to be a significant concern. The evolution of these attacks spans from traditional SQL injection and command injection to modern prompt injection in AI systems [10]. The underlying principle remains the same: introducing untrusted data into a trusted application's execution flow. Prompt injection, in particular, leverages the natural language interface of LLMs to manipulate their behavior, leading to unauthorized actions or data exfiltration [7][11].

Chain Reactions and Amplification

A critical aspect of modern exploitation is the concept of attack chaining, where multiple, seemingly minor vulnerabilities are linked together to achieve a significant impact. This can involve escalating an "out-of-scope" Cross-Site Scripting (XSS) vulnerability into a full account takeover through techniques like iframe sandwiching [12]. Similarly, arbitrary file writes (AFW), while often demonstrating limited impact in isolation, can be escalated to Remote Code Execution (RCE) by finding an appropriately executed destination within the target environment [13]. The success of these chains depends heavily on understanding the execution context, including the operating system, language runtime, and framework versions [13].

Exploiting Trust and Authentication Mechanisms

Attacks frequently target trust boundaries and authentication mechanisms. For example, misconfigured CORS policies or weak WAF rules can facilitate unauthorized access or data leakage [1]. Bypassing multi-factor authentication (MFA) in AI agents and Interactive Voice Responses (IVRs) is another area of concern, often achieved through clever social engineering or exploiting flaws in the authentication logic [7]. The eIDAS qualified signature system, used by numerous banks and government agencies, was found to have vulnerabilities allowing websites to read sensitive card data and potentially achieve drive-by RCE by leveraging flaws in its signing extension and native host [14].

Kernel-Level Exploitation and Evasion

For privilege escalation, attackers target kernel vulnerabilities. Techniques involve modifying kernel memory structures like the SSDT (System Service Dispatch Table) or using kernel drivers to gain SYSTEM privileges, even on systems with advanced memory isolation and kernel protection features [15]. Advanced techniques also focus on EDR evasion, such as decrypting malicious code in GPU memory or weaponizing Windows Cloud Filter for fileless execution [15].

Supply Chain Compromise

The increasing reliance on third-party code and dependencies creates significant supply chain risks. Vulnerabilities in package managers like RubyGems or build systems like GitHub Actions can be exploited to compromise entire software supply chains [4]. This includes issues like cache poisoning, where unauthenticated attackers can retrieve sensitive API keys from a CDN, or attacks that hijack cloud identities through reclaimed namespaces in global OIDC issuers [4][16].

Notable Techniques and Emerging Attack Surfaces

The application security community continuously innovates in discovering and weaponizing new vulnerabilities. Several recent developments and techniques stand out for their novelty and potential impact.

AI-Driven Vulnerability Discovery and Exploitation

AI, particularly LLMs, are no longer just tools for defense; they are becoming potent weapons for offense. Researchers are using AI to discover vulnerabilities that would be time-consuming or impossible for humans to find. For instance, an AI tool identified a 20-year-old RCE vulnerability in the original Call of Duty game [2]. More sophisticated applications include using LLMs to chain together complex exploit primitives, such as identifying a pre-authentication RCE in WordPress by combining batch API validation bugs, SQL injection, cache poisoning, and privilege escalation [4]. The "HTTP Terminator" project demonstrates how AI can autonomously invent new attack techniques and exploit live websites at scale, including novel HTTP desync triggers and exploits against banks, security solutions, and government infrastructure [5].

Weaponizing AI Agents and Chatbots

Beyond finding vulnerabilities in traditional applications, AI systems themselves are becoming targets. Attackers can weaponize chatbots via email to send phishing messages or trick them into performing unauthorized actions [7]. Prompt injection is a prominent technique here, where carefully crafted inputs can manipulate LLMs into revealing sensitive information, bypassing safety guardrails, or executing malicious commands. The OWASP LLM Top 10 project identifies categories of vulnerabilities specific to LLM applications, providing a framework for understanding these new risks [11][17][18]. Hijacking AI coding agents from major providers like Anthropic, Google, and OpenAI through a single untrusted GitHub issue is a stark reminder of the potential for compromising development workflows [6].

Exploiting Browser and Web Technologies

Novel attacks continue to emerge targeting web browsers and their extensions. CSS can be weaponized within webmail clients to bypass sanitizers, exfiltrate tokens, spoof UI elements, and steal passwords across various email providers [4]. Techniques include nesting attribute selectors to brute-force tokens or using font-height oracles with animations to exfiltrate numeric tokens when Content Security Policy (CSP) blocks external resources. Understanding the attack surface of browser extensions, including content scripts, service workers, and extension pages, is crucial, as misconfigurations or vulnerabilities in these components can lead to significant compromise [19].

Arbitrary File Write Escalation in Modern Environments

The path from an arbitrary file write (AFW) to RCE is often challenging, particularly in modern, stripped-down containerized environments [13]. Research categorizes AFW-to-RCE techniques by exploiting the operating system, language runtime, or framework. While traditional methods like cron or SSH access are less common in containers, techniques like writing to /etc/ld.so.preload remain potent as they trigger on every new process execution. The "Write Once, Shell Everywhere" research systematically catalogues existing techniques, introduces environment fingerprinting to turn writes into read primitives, and provides specific methods for containerized environments [13].

Exploiting Connected Systems and IoT

Vulnerabilities in systems that bridge the physical and digital worlds are increasingly reported. The Connective signing extension, used by major Belgian banks and government agencies, was found to allow unauthorized access to eID and Maestro card data and enabled drive-by RCE through a vulnerable native binary [14]. This highlights the critical need for security in applications that interact with hardware and sensitive personal identity data. Hacking Android and IoT applications by example is a focused area of study, emphasizing deep dives into static and dynamic analysis, instrumentation techniques, and specific exploit methods [20].

Detection and Prevention Strategies

Mitigating these evolving threats requires a multi-layered approach, integrating proactive security measures into the entire SDLC and leveraging intelligent detection capabilities.

Shift-Left Security and Developer Enablement

A fundamental shift in security strategy is moving security left in the SDLC. This involves empowering developers with the tools and knowledge to identify and fix vulnerabilities early in the development process [9]. Tools that provide immediate feedback and analysis, such as static analysis security testing (SAST) and dynamic analysis security testing (DAST) solutions, are crucial. A developer-first mindset, focusing on making security tasks easier and less burdensome for developers, can foster better adoption and a stronger security culture [9].

Threat Modeling for Complex Systems and AI

Traditional threat modeling approaches need to be adapted for modern, complex systems and especially for AI agents and workflows [21]. For AI agents, this involves understanding agent architecture, tool invocation, and the unique trust boundaries that emerge when LLMs interact with external services and data. Threat modeling AI-driven systems requires identifying assets like the model, tool APIs, and data stores, and analyzing threats such as prompt manipulation, unauthorized tool use, and data leakage. Story-driven threat modeling can be applied to AI workflows by deriving abuse cases from user prompts and agent goals [21].

Runtime Protection and EDR Evasion Techniques

For detection and prevention at runtime, understanding sophisticated evasion techniques is vital. This includes techniques that target kernel-level protections [15] or employ fileless execution methods. Researchers are developing advanced methods to bypass EDR solutions by leveraging capabilities like GPU memory for decryption or using Windows Cloud Filter mechanisms [15]. For application-level runtime protection, securing AI agents against prompt injection, excessive agency, and tool misuse is paramount [21]. This often involves implementing robust prompt controls, input validation, and principle of least privilege for tool usage.

Secure AI Development and Governance

As AI becomes integral to development, securing its use is critical. This involves establishing governance for AI tools, including addressing the risks of "Shadow AI" [8]. For AI agents, security best practices include:

Supply Chain Security Measures

Addressing supply chain risks requires a multi-faceted approach. This includes securing build pipelines, vetting dependencies, and implementing stricter controls on package management systems [4]. For GitHub Actions, safer defaults for pull_request_target and read-only caches for untrusted triggers are key mitigations. For package repositories, ensuring proper cache control and authentication mechanisms are in place is essential to prevent API key exfiltration [4].

Tooling and Ecosystem

The application security tooling landscape is rapidly evolving, with AI and automation playing an increasingly significant role.

AI-Powered Security Tools

AI is rapidly being integrated into security tooling for various purposes, including vulnerability discovery, code review, and threat intelligence. Tools like XBOW demonstrate autonomous penetration testing capabilities, discovering dozens of vulnerabilities and producing proof-of-concept exploits in real-time [22]. AI assistants are also being used to automate tasks like summarization of investigations and translation of detection rules within SOC environments [8]. The development of AI gateways for AI services, focused on prompt protection, is also emerging [23].

Enhanced Debugging and Analysis Tools

For web application security, tools like Burp Suite remain foundational. Resources like the "Testing Handbook - Burp" provide in-depth guidance on leveraging its capabilities [24]. For browser extension security, understanding how to debug them by accessing developer tools and enabling specific settings is crucial for identifying vulnerabilities [19].

Specialized Tooling for Modern Environments

For cloud-native environments, tools and techniques for securing Kubernetes clusters are essential. This includes implementing admission controller policies, building secure container images, and understanding the capabilities and limitations of eBPF for security monitoring [25]. The use of tools for AI Red Teaming, such as those found in the "GenAI Security Training" repository, provides hands-on labs and practical demonstrations of adversarial techniques against AI systems [17].

Frameworks for AI Security Research

The growing complexity of AI security has led to the development of frameworks and projects aimed at standardizing security practices. The OWASP GenAI Security Project is developing resources and guidance for securing LLMs and autonomous agents [18]. The "AI Agent Security Masterclass" covers attacking and defending autonomous AI systems, utilizing concepts like the Model Context Protocol (MCP) and Retrieval-Augmented Generation (RAG) [21].

Recent Developments and Future Trends

The rapid advancements in AI and the evolving threat landscape point towards several key trends shaping the future of application security.

The "Wild West" of AI Security

The integration of AI into cybersecurity is still in its nascent stages, leading to a period often described as the "Wild West" of AI security [8]. Security teams are actively grappling with understanding agent-to-agent interactions, enterprise AI governance, and new attack surfaces introduced by AI technologies [8]. The rapid adoption of AI tools is outpacing the development of understanding and security controls, creating a significant challenge for organizations [8].

AI as an Autonomous Research Partner

The capability of AI to perform novel security research is a significant development. The "HTTP Terminator" project showcases AI autonomously inventing new attack techniques and weaponizing them against live targets [5]. This suggests a future where AI acts not just as a tool but as a research partner, accelerating the discovery of new vulnerabilities and attack methods. This trend is also evident in CTF scenarios where LLMs are automating challenge resolution, forcing a re-evaluation of challenge design and competition dynamics [3].

The Rise of Agentic AI and its Security Implications

Agentic AI, where AI systems act autonomously to perform tasks, is a major trend. This introduces new security considerations related to delegated access, agent-to-agent interactions, and the potential for AI agents to make irreversible changes or propagate errors [8][21]. Securing these systems involves understanding concepts like the Model Context Protocol (MCP) and developing robust defense mechanisms against prompt injection, excessive agency, and tool misuse [21].

Focus on Supply Chain and Software Composition Analysis (SCA)

With the increasing prevalence of software supply chain attacks, there's a heightened focus on securing dependencies and the development pipeline [4][9]. Efforts are underway to enhance the resilience of SCA tools against techniques like container image obfuscation and to improve SBOM generation algorithms [23]. The security of AI-generated code also presents a new facet of supply chain risk that needs careful management [9].

Continued Importance of Foundational AppSec Skills

Despite the rapid evolution of technologies like AI, foundational application security principles remain critical. Skills in understanding injection attacks, exploit chaining, and leveraging tools like Burp Suite are still indispensable [10][12][13][24]. The ability to perform thorough reconnaissance, understand protocol nuances, and conduct in-depth analysis is crucial for both offensive and defensive security professionals.

Where to Go Deeper

For practitioners looking to deepen their expertise in application security, several avenues offer continuous learning and community engagement.

Conference Talks and Recordings

Major security conferences are invaluable resources for staying abreast of the latest research and techniques. Key events like Black Hat USA, DEF CON, and OWASP Global AppSec conferences consistently feature cutting-edge talks on application security, AI security, and emerging threats [1][10][12][6][4][5][25][9][23][26][27][28][29][30][31][32][33][34][35][22][36][37][38][18][39][40][41][42][43][44][45][46][47][48][49][50][51][52][53][54][21][20][55][56][57][58][59][60]. Many of these conferences make their talk recordings available online, offering a wealth of knowledge accessible remotely [15][28][35][61][58][62].

Online Resources and Training

Community Engagement

Participating in CTFs (Capture the Flag competitions) offers a practical way to hone skills and apply learned techniques in a gamified environment [3][32][39][43][20]. Engaging with the broader security community through forums, mailing lists, and social media can provide valuable insights, mentorship, and networking opportunities. Following prominent researchers and organizations on platforms like LinkedIn can also offer a steady stream of expert commentary and research findings [68][69].

Academic and Research Papers

For those seeking in-depth, peer-reviewed research, IEEE Symposium on Security and Privacy is a premier venue [47]. Academic papers often present novel theoretical advancements and empirical studies that can inform future defensive strategies and identify emerging attack vectors.

Sources cited in this guide

  1. Chain Reaction: From Isolated Vulnerabilities to Full System Compromise (RomHack Camp 2026) — cfp.romhack.io
  2. Pwning Call of Duty 1: a 20-year-old RCE, found in an evening with AI — zolder.io
  3. CTFs in the AI Era — blog.includesecurity.com
  4. [tl;dr sec] #341 - Hugging Face Incident Black Hat Talk, CSS Bomb in your Inbox, GitHub Supply Chain Security Improvements — tldrsec.com
  5. Can AI do novel security research? Meet the HTTP Terminator — portswigger.net
  6. AMA with Black Hat Speakers Lidor B. & Elad Meged (Pre-Auth RCE in Enterprise Java, Hijacking AI Coding Agents) — pwnhackers.substack.com
  7. Hacking AI customer service agents — intigriti.com
  8. 5 Takeaways from Black Hat x DEF CON 2025 — cybersecuritypulse.net
  9. Securing next-gen development: Lessons from Trust Bank and TASConnect — snyk.io
  10. Injection Attacks Through the Ages: From Turing Machines to Prompt Injection — developerstv.com
  11. Jackpot: a browser lab of 10 deliberately vulnerable LLM apps, one per OWASP LLM Top 10 category — hego.red
  12. From Out of Scope XSS to Account Takeover: The Iframe Sandwich (OWASP talk) — tsecurity.de
  13. Write Once, Shell Everywhere - Turning Arbitrary File Writes into RCE (DEF CON Bug Bounty Village) — ethiack.com
  14. DEF CON talk: 8 in 10 Banks in Belgium HATE This One Weird eID RCE — amibeingpwned.com
  15. Conference recordings are now online: Attack Techniques & Methods — dbu.gs
  16. [tl;dr sec] #332 - I've Joined OpenAI, fwd:cloudsec, AWS Well Architected Supply Chain Security — tldrsec.com
  17. GitHub - schwartz1375/genai-security-training — github.com
  18. OWASP Global AppSec EU 2025 - GenAI Focus — genai.owasp.org
  19. [HackerNotes Ep.95 & Ep.96] Cookies, Caching & Attacking Chrome Extensions with MatanBer — blog.criticalthinkingpodcast.io
  20. Hacking Android and IOT Apps by Example - DEF CON Training LV 2026 — training.defcon.org
  21. AI Agent Security Masterclass: Attacking and Defending Autonomous AI Systems - DEF CON Training — training.defcon.org
  22. XBOW at Black Hat & DEF CON: AI Agents for Offensive Security — xbow.com
  23. Top security talks from KubeCon Europe 2025 — wiz.io
  24. Testing Handbook - Burp — docs.google.com
  25. Top security talks from KubeCon Europe 2024 — wiz.io
  26. Black Hat USA 2025 Briefings Schedule — blackhat.com
  27. Black Hat USA 2025 - YouTube Playlist — youtube.com
  28. Black Hat Official YouTube Channel — youtube.com
  29. DEF CON 33 AppSec Village — appsecvillage.com
  30. DEF CON 33 Aerospace Village Talk Schedule — aerospacevillage.org
  31. About NDC Security 2026 — ndcsecurity.com
  32. DEF CON 33 Hacking Conference 2025 — USF — usf.edu
  33. DEF CON 33 (2025) — Security.World — security.world
  34. What to Expect from BSides, Black Hat, and DEF CON 2025 — theregister.com
  35. DEFCON Conference — Official YouTube — youtube.com
  36. Black Hat USA 2025 Briefings Schedule — blackhat.com
  37. Approov Events and Conferences — approov.io
  38. OWASP Global AppSec USA 2025 - CFP — sessionize.com
  39. OWASP Global AppSec EU 2025 (Barcelona) — owasp.glueup.com
  40. OWASP Global AppSec USA 2025 (Washington DC) — genai.owasp.org
  41. OWASP Global & Regional Events — owasp.org
  42. OWASP AppSec Days Developer Security Summit — owasp.org
  43. OWASP 25th Anniversary Virtual Conference — sessionize.com
  44. AppSec & Cybersecurity Events Calendar 2026: 60+ Conferences — derscanner.com
  45. Annual Computer Security Applications Conference — acsac.org
  46. GPSEC Cybersecurity Conference — guidepointsecurity.com
  47. IEEE Symposium on Security and Privacy 2026 — sp2026.ieee-security.org
  48. The Elephant in AppSec Conference — theelephantinappsec.com
  49. RSA Conference 2026 — rsaconference.com
  50. Cybersecurity Conferences 2026-2027 — infosec-conferences.com
  51. NDC Security 2026 — ndcsecurity.com
  52. The Best Security Conferences & Events 2026 — splunk.com
  53. LASCON – Lonestar Application Security Conference — lascon.org
  54. HTB COAE: Introducing the new standard for AI Red Teaming — youtube.com
  55. Black Hat USA 2026 Training Schedule — blackhat.com
  56. Black Hat Briefings - Wikipedia — en.wikipedia.org
  57. Introducing the OWASP Nettacker Project - Speaker Deck — speakerdeck.com
  58. Global AppSec San Francisco 2024 — youtube.com
  59. The Scariest Things We Saw at Black Hat 2021 — medium.com
  60. CircleCityCon 8.0: Happy Little Conference — registration.circlecitycon.com
  61. (96) NahamSec - YouTube — youtube.com
  62. DEFCONConference — youtube.com
  63. BSidesSLC 2026 — sessionize.com
  64. OWASP Videos — videos.owasp.org
  65. OWASP Videos — videos.owasp.org
  66. GitHub - schwartz1375/genai-essentials — github.com
  67. Web App Pentesting in the AI Era — blog.includesecurity.com
  68. Marcus Hutchinsthe cybersecurity researcher best known for stopping the WannaCry ransomware attackoffers his perspectives on Mythos | The Cyber Security Hub — linkedin.com
  69. #burp #pentest #ai #hackerassociate #cybersecurity #infosec… | Harshad Shah — linkedin.com
📚 This guide is synthesized from the full text of resources curated in the Talks library, and refreshed as new material is added.