2.0 -//Pentabarf//Schedule//EN PUBLISH K7JNE8@@cfp.bsidesvienna.at -K7JNE8 MCP - Most Concerning Protocol en en 20260627T093000 20260627T101500 0.04500 MCP - Most Concerning Protocol Basically our talk shows all the ways attackers can leverage MCP for RCE and persistence. In this talk we demonstrate “MCP hijacking in the wild” through an attack demo that shows how a compromised or malicious MCP tool execution path can become an attacker control channel. Critically, we show why common hardening approaches are insufficient in practice by demonstrating bypass and/or time-of-check/time-of-use gaps during the demo itself. We close with a proof-of-concept integrity protection method designed to raise the bar against MCP toolchain compromise by enforcing trustworthy tool identity and invocation integrity, with practical guidance PUBLIC CONFIRMED Medium Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/K7JNE8/ Mittlerer Saal (Track 1) Paul Zenker PUBLISH LQVWHG@@cfp.bsidesvienna.at -LQVWHG Hey Claude, find 0days - Using AI for Vulnerability Research & Red Teaming en en 20260627T102000 20260627T110500 0.04500 Hey Claude, find 0days - Using AI for Vulnerability Research & Red Teaming We'll examine the capabilities of AI itself, how we can integrate it into our research to find more bugs more quickly, and how well it can be used for weaponization. To visualize this, we'll walk through at least one real-world vulnerability: my original approach, how that approach can be improved by AI, and to what degree. But the advantages of AI come at a (hidden) cost: we're sharing everything with a third party, be it Anthropic, Google, or whoever. Would you upload your C2 beacon to VirusTotal during a Red Team? So, where exactly should we draw the line on AI? PUBLIC CONFIRMED Medium Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/LQVWHG/ Mittlerer Saal (Track 1) Niels Pfau PUBLISH C37PNQ@@cfp.bsidesvienna.at -C37PNQ BOFs in the Background: Async object file execution in modern C2 frameworks en en 20260627T111000 20260627T121000 1.00000 BOFs in the Background: Async object file execution in modern C2 frameworks In [last year's](https://cfp.bsidesvienna.at/bsidesvienna-0x7e9-2025/talk/VCQATE/) talk, I introduced Conquest, a command and control framework written in Nim. Since then, the project has been under heavy development and has seen countless new features, fixes and improvements. One of the most substantial changes was the addition of a Python module system, which allows red teamers and penetration testers to turn their favourite BOFs into Conquest commands. This drastically decreases the amount of functionality that has to be included in the base agent, because post-exploitation capabilities can simply be executed as external BOFs. However, regular BOF loading is boring. There are countless open-source implementations and it's a feature in almost all C2 frameworks on the market. On the other hand, the number of frameworks that support asynchronous BOFs can almost be counted on one hand. In fact, Conquest is one of the only C2s that features stable async object file execution, even when the agent memory is obfuscated or encrypted, something that until now has been exclusive to commercial tooling. This talk goes beyond proof-of-concepts by also featuring novel use-cases for async BOFs that benefit real engagements. - [TGT Monitoring](https://github.com/jakobfriedl/tgt-monitor-bof) - [KeePass Abuse](https://github.com/jakobfriedl/keelog-bof) - [Clipboard Monitoring](https://github.com/jakobfriedl/clipboard-monitor-bof) - ... PUBLIC CONFIRMED Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/C37PNQ/ Mittlerer Saal (Track 1) Jakob Friedl PUBLISH TVJ7J8@@cfp.bsidesvienna.at -TVJ7J8 Net Shredder: Coverage-Guided Network Fuzzing for the Linux Kernel en en 20260627T131500 20260627T140000 0.04500 Net Shredder: Coverage-Guided Network Fuzzing for the Linux Kernel Net Shredder was built out of desire to build the simplest fuzzing setup possible for the Linux kernel: - Unlike grammar based fuzzing, this approach doesn’t require rewriting the protocol spec in terms of fuzzer grammar; - Just a basic setup and a fuzzing corpus is enough; - Only small target modifications required. This talks covers building it as a solo project I did in 2025: - Collecting remote KCOV outside process context, turning raw PCs into edge coverage - The struggle of restricting coverage to the targeted subsystem (2 failed and the one that worked) - Three bugs it found, including CVE-2025-22037 with bug analysis. PUBLIC CONFIRMED Medium Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/TVJ7J8/ Mittlerer Saal (Track 1) Vyacheslav "Slava" Moskvin PUBLISH W9MWHG@@cfp.bsidesvienna.at -W9MWHG Project TEMPA: Getting Phon(e)key with Tesla Security Again en en 20260627T140500 20260627T145000 0.04500 Project TEMPA: Getting Phon(e)key with Tesla Security Again PUBLIC CONFIRMED Medium Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/W9MWHG/ Mittlerer Saal (Track 1) Martin Herfurt PUBLISH J9QR9K@@cfp.bsidesvienna.at -J9QR9K Back to the Binary: Revisiting Similarities of Android Apps en en 20260627T152000 20260627T160500 0.04500 Back to the Binary: Revisiting Similarities of Android Apps Quickly determining whether two apps are related is interesting for a couple of reasons: Malware detection, impersonification, lineage analysis, and more. But making this decision based on their code has many pitfalls, even if we don't take obfuscation into account. Stale tools, no updates, missing dependencies - instead of those we'll use the binary representation of apps and a combination of open source tools to get an estimator for code similarity. The talk will cover some Android internals based on examples, and will provide a practical blueprint how to calculate code-based similarity of apps based on open-source tools. Suitable for non-reverse-engineers. PUBLIC CONFIRMED Medium Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/J9QR9K/ Mittlerer Saal (Track 1) Jakob Bleier PUBLISH UDUZHL@@cfp.bsidesvienna.at -UDUZHL Windows CE Memory Archaeology - Recovering Files from Windows CE 5.0 on NXP i.MX28 NAND Flash en en 20260627T161000 20260627T170500 0.05500 Windows CE Memory Archaeology - Recovering Files from Windows CE 5.0 on NXP i.MX28 NAND Flash One of the first steps in embedded systems vulnerability research is to access the firmware, obtained from firmware updates or from a physical device. How hard can it be? Usually not that much. But in this case, after failing miserably using the quick and easy standard approach, we doubled down and layer by layer uncovered how files are stored by Windows CE 5.0 on devices using the NXP i.MX28 processor. The single topics are nothing new, but due to the age and targeting (mobile phones and PDAs) of most Windows CE tooling and hardware specifics, ready-to-use tools failed repeatedly. Starting with no clue about Windows CE internals, our approach was to leverage a combination of ancient technology (Windows XP) and modern technology (AI) to understand what we are dealing with and to adapt and build our customized tooling. PUBLIC CONFIRMED Medium Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/UDUZHL/ Mittlerer Saal (Track 1) Gerhard Hechenberger PUBLISH FHQ7JH@@cfp.bsidesvienna.at -FHQ7JH Zero Files, Zero Noise: Checkmate in Three. en en 20260627T171000 20260627T181000 1.00000 Zero Files, Zero Noise: Checkmate in Three. Modern intrusions rarely fail because one alert says everything. They fail when small, imperfect traces are recognized early enough and correlated into a timeline the operator did not expect. In this case, OSINT and TTP comparison helped us put the observed behavior into context. We found overlaps with previously reported activity dating back to 2022, showing which parts of the operator’s tradecraft remained stable and which had been adapted to the current defensive environment. This makes the case relevant now: the tooling changed, but the operational habits, decision points, and forensic residue remained recognizable. The focus is not attribution for its own sake, but understanding how recognizable patterns can support an active investigation. We will show how the investigation moved through three decisive phases: identifying the initial execution pattern, separating real attacker activity from low-signal security telemetry, and reconstructing the operator’s movement from endpoint, identity, and network evidence. Along the way, we will discuss what the defender’s environment made possible, where visibility was missing, and why the attacker’s interaction with Defender became an investigative advantage rather than just a detection failure. The focus is Incident Response depth: not attribution, not tool training, but the reconstruction of attacker behavior from imperfect telemetry. Attendees will see how current fileless and defense-evasion techniques can still produce useful traces, which log sources made the difference in this case, and how two initially unclear Defender for Endpoint notifications were turned into a defensible incident timeline. PUBLIC CONFIRMED Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/FHQ7JH/ Mittlerer Saal (Track 1) Jonas Plitt PUBLISH AJN9TG@@cfp.bsidesvienna.at -AJN9TG It's Not You, It's Your Dependencies: A Nerdy Lawyer's Guide to the Software Supply Chain en en 20260627T100500 20260627T103500 0.03000 It's Not You, It's Your Dependencies: A Nerdy Lawyer's Guide to the Software Supply Chain PUBLIC CONFIRMED Short Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/AJN9TG/ Dachsaal (Track 2 ) Katharina Bisset PUBLISH Z7XYAA@@cfp.bsidesvienna.at -Z7XYAA Locks opened keyless the hard way en en 20260627T104000 20260627T114000 1.00000 Locks opened keyless the hard way A lock can be a hard challenge to open without a key. Regularly these are installed in a door. We do it a little bit easier, we do not install them, we hold them in our own hand. But additionally we do it a little bit harder, because we do not want to destroy them. If you want to open the door, there are much easier ways, but they will damage something. You do not need to know, how a lock works, to pick them. Usually, we do not explain it, we let beginners open the lock for the first time and anybody is surprised, it works easy. But it was an easy lock. But what are the differences between an easy and a hard lock? So for that we have to explain a lock and the key. Which technical behaviors are used to make it more and more challenging. Somebody could think, the manufactures do it for more fun for the lock pickers. For more fun, it is nice to use rules for competitions. The belt-system is one. Anybody can participate. And you can use it to describe the fun - sorry - the easiness or hardness of a lock for lock pickers. PUBLIC CONFIRMED Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/Z7XYAA/ Dachsaal (Track 2 ) deac TiborElias PUBLISH 3GKCFS@@cfp.bsidesvienna.at -3GKCFS Your Firewall Won’t Save You From a Crowbar: Introducing the PACT Framework en en 20260627T114500 20260627T120500 0.02000 Your Firewall Won’t Save You From a Crowbar: Introducing the PACT Framework We spend millions hardening the cloud, fine-tuning EDRs, and chasing zero-days, yet we consistently forget one glaring reality: an adversary can just walk through the front door with a high-vis vest and a clipboard. For years, the cybersecurity community has worshiped at the altar of the MITRE ATT&CK® framework to track digital adversary behaviors. Meanwhile, physical security has remained trapped in a legacy bubble of "locks, blocks, and glocks," utilizing entirely different risk languages. When a threat actor breaks into a facility to deploy a network implant, is it a cyber attack or a physical breach? It's both-and our current defense matrices fail to map this convergence. Enter PACT (Physical Access & Control Taxonomy). Built to bridge the grand canyon between meatspace and cyberspace, PACT is an open-source, community-driven framework that translates physical tactics, techniques, and procedures (TTPs) into a structured matrix that mirrors the MITRE ATT&CK format. This talk will rip off the band-aid of siloed security, break down the anatomy of the PACT matrix, and demonstrate how Red and Blue teams can finally use a single, unified language to hunt threats across both the digital and physical realms. PUBLIC CONFIRMED Short Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/3GKCFS/ Dachsaal (Track 2 ) Darius Beckert PUBLISH MAKHRG@@cfp.bsidesvienna.at -MAKHRG "The Human Factor. Cybersecurity's weakest link or most adaptive defense?" en en 20260627T130500 20260627T135500 0.05000 "The Human Factor. Cybersecurity's weakest link or most adaptive defense?" In our talk we scrutinize the critical role of human behavior in cyber defense and how it reframes defense strategies beyond purely technical or machine-driven approaches. Human cognition acts as complementary system, not just a weak link. This creates a hybrid model, where adaptive human-decision making works in tandem with machine efficiency. One of its core strengths lies in pattern recognition and contextual understanding. Security analysts are often able to spot unusual behaviors or subtle anomalies that automated systems might overlook. Equally important are creativity and intuition, which allow humans to anticipate novel attack methods and think beyond established rules or signatures. Like the immune system, humans are capable of adaptability, adjusting strategies quickly in response to emerging threats. In addition, they bring ethical judgment and strategic decision-making, enabling a nuanced evaluation of risks that goes beyond metrics and algorithms. However, the human factor also introduces vulnerabilities. Alert fatigue and cognitive overload can reduce vigilance, leading to missed threats. Agenda: Assumptions „The problem is sitting in front of the screen“ (User) Security Behavior in Organizations – Security Compliance vs. Security Participation People don’t fail randomly. They fail predictably within the system we design. Organizational, Individual & Environmental influencing factors & dimensions of security behavior and the question “Why does security often not work in practice” „Policies are made to pass audits – not actually for you to follow“ Practical examples: force of habits, alert fatigue, psychological Acceptability: Phishing “Computers are weird” Key Takeaways PUBLIC CONFIRMED Medium Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/MAKHRG/ Dachsaal (Track 2 ) Yvonne Bauer Wolfgang Ettlinger PUBLISH USUBHG@@cfp.bsidesvienna.at -USUBHG The sky is no longer the limit en en 20260627T140000 20260627T144500 0.04500 The sky is no longer the limit While drones offer immense potential, they have introduced a complex new landscape of risk. This talk examines the "Gray Rhino" threats of 2026 high-probability security challenges often ignored until impact. We will analyze the transition of advanced drone tech from warzones to criminal enterprises, including the rise of unjammable fiber-optic controls and AI-enabled targeting. Beyond physical safety, we’ll explore the escalating vulnerabilities of critical infrastructure and the too slow shifting regulatory response to aerial sabotage. We will dive deeper in the actual risk and legal mitigations we could and should take. The are shaping the battlefield already so society will be next. PUBLIC CONFIRMED Medium Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/USUBHG/ Dachsaal (Track 2 ) Stephan Van Dyck PUBLISH VDQMNY@@cfp.bsidesvienna.at -VDQMNY Defending Identity Infrastructure of the Active Directory with Deception Technologies en en 20260627T153000 20260627T160000 0.03000 Defending Identity Infrastructure of the Active Directory with Deception Technologies Modern enterprise and governmental IT infrastructures rely heavily on identity systems to control access to critical resources. Among these systems, Active Directory (AD) remains the dominant identity management platform and therefore represents one of the most attractive targets for cyber attackers. Once an adversary gains an initial foothold inside a network—whether through phishing, credential compromise, or exploitation of a vulnerable system—the next strategic objective is typically the compromise of the identity infrastructure. By targeting Active Directory, attackers can escalate privileges, move laterally across systems, establish persistence, and ultimately achieve full domain dominance. In recent years, numerous large-scale breaches have demonstrated that attackers often operate inside networks for extended periods before detection. During this dwell time, adversaries perform extensive reconnaissance activities within the directory environment. These reconnaissance steps typically involve directory enumeration, credential harvesting, privilege path analysis, and the identification of high-value targets such as administrative accounts or privileged service identities. Tools such as BloodHound are commonly used by attackers to map relationships between users, groups, and systems, enabling them to identify potential privilege escalation paths. Traditional detection approaches frequently struggle to identify these early-stage reconnaissance activities. Security monitoring systems generate vast volumes of log data, making it difficult to distinguish malicious behavior from legitimate administrative activity. As a result, attackers are often detected only after significant damage has already occurred, such as during lateral movement, privilege escalation, or the deployment of ransomware. This presentation explores a proactive defensive strategy based on deception technologies, specifically focusing on the use of Active Directory honeypots and canary tokens as early-warning mechanisms within enterprise identity infrastructures. Deception techniques allow defenders to strategically place decoy assets within the directory environment that appear legitimate to attackers but are monitored closely by security teams. These assets function as highly sensitive tripwires: any interaction with them can immediately indicate suspicious or malicious activity. The talk will examine how deception-based controls can be integrated directly into Active Directory environments through multiple layers of defensive instrumentation. These layers may include honey user accounts, decoy service identities, embedded credential lures, deceptive group memberships, and carefully crafted scripts placed in shared directories such as SYSVOL. When adversaries enumerate the directory or attempt to leverage these artifacts during their reconnaissance phase, detection mechanisms can trigger alerts and provide defenders with early indicators of compromise. Particular attention will be given to the implementation of canary tokens, a form of digital tripwire that generates alerts when accessed or used. By embedding tokens within configuration files, service account descriptions, administrative scripts, or backup documentation, defenders can create realistic artifacts that attackers are likely to interact with during credential discovery or environment mapping. When these tokens are triggered, alerts can be delivered through email notifications, security monitoring platforms, or integrated SIEM systems, providing immediate visibility into suspicious activity. The presentation will also demonstrate how deception mechanisms can be incorporated into Purple Teaming methodologies, enabling collaboration between offensive and defensive security teams. Through controlled attack simulations performed by Red Teams, organizations can test whether their defensive monitoring systems successfully detect interactions with deception artifacts. Blue Teams can then refine detection rules, improve security monitoring workflows, and strengthen incident response procedures based on these insights. Beyond detection, deception technologies offer substantial benefits for digital forensic investigations. When attackers interact with honeypot objects or canary tokens, investigators can collect valuable forensic artifacts, including authentication attempts, directory queries, network activity, and timestamps associated with attacker behavior. These artifacts allow security teams to reconstruct the attacker’s timeline, understand the reconnaissance techniques used, and determine the scope of a potential compromise. In this way, deception systems act not only as detection mechanisms but also as forensic sensors that capture high-fidelity evidence of adversary activity. A practical implementation scenario will be presented in which an attacker performs Active Directory reconnaissance using common tools such as BloodHound. During the enumeration process, the attacker encounters a realistic honey account that appears to belong to a backup service. Embedded credential lures and a hidden token trigger lead the attacker to interact with a decoy artifact. As soon as the attacker attempts to use the discovered credentials or access the embedded token, an alert is generated through a monitoring system such as Microsoft Defender for Identity or a SIEM platform. This early detection enables defenders to begin incident response activities before the attacker can escalate privileges or achieve domain control. In addition to demonstrating the technical implementation of deception techniques, the presentation will discuss operational considerations and potential challenges. Effective deception requires realistic design to avoid detection by skilled adversaries. Honey accounts must follow believable naming conventions, deceptive artifacts must appear consistent with normal operational practices, and monitoring infrastructure must be properly configured to ensure reliable alerting. Maintenance and lifecycle management of deception assets are also important to maintain credibility within the environment. The session concludes by exploring the broader strategic implications of deception technologies in modern cyber defense architectures. As attackers increasingly target identity systems, defensive strategies must evolve to detect adversaries earlier in the attack lifecycle. Deception-based detection mechanisms provide a powerful complementary layer to traditional monitoring controls by focusing on attacker behavior rather than relying solely on anomaly detection within large datasets. By integrating honeypots, canary tokens, Purple Teaming exercises, and digital forensic analysis, organizations can significantly improve their ability to detect intrusions at an early stage, reduce attacker dwell time, and collect valuable intelligence about adversary techniques. The insights gained from these systems can help security teams strengthen defensive capabilities, refine detection strategies, and ultimately improve the resilience of enterprise identity infrastructures against modern cyber threats. PUBLIC CONFIRMED Short Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/VDQMNY/ Dachsaal (Track 2 ) Ahmed Hassan PUBLISH MZYRYC@@cfp.bsidesvienna.at -MZYRYC Quantum Secure Communication en en 20260627T160500 20260627T165000 0.04500 Quantum Secure Communication PUBLIC CONFIRMED Short Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/MZYRYC/ Dachsaal (Track 2 ) Andreas Neuhold PUBLISH H7CXVP@@cfp.bsidesvienna.at -H7CXVP Attackbed: A Damn Vulnerable Network for Profit and Fun en en 20260627T165500 20260627T172500 0.03000 Attackbed: A Damn Vulnerable Network for Profit and Fun PUBLIC CONFIRMED Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/H7CXVP/ Dachsaal (Track 2 ) Wolfgang Hotwagner PUBLISH AXV9VA@@cfp.bsidesvienna.at -AXV9VA When Machines Hack Back: How AI Rewrote the Threat Landscape in 12 Months en en 20260627T094500 20260627T101500 0.03000 When Machines Hack Back: How AI Rewrote the Threat Landscape in 12 Months In 2025, cybercrime became a continuous, machine-speed system. Exploitation attempts surged 25% to 120 billion globally, time-to-exploit dropped to 24–48 hours, and AI tools like WormGPT eliminated the skill barrier for attackers. Identity replaced malware as the primary attack vector, with billions of stolen credentials enabling login-based intrusions at scale. Deepfakes now target executives directly for fraud and access. The Claude Mythos announcement in April 2026 marked an inflection point: autonomous discovery of decade-old zero-days across all major operating systems overnight, for under $2,000 per exploit. This talk covers current threat data, the implications for defenders, and why Continuous Threat Exposure Management and AI-enabled defense are no longer optional. PUBLIC CONFIRMED Short Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/AXV9VA/ Kreativraum 3.1 (Track 3 - Women4Cyber/Rookie) Ronke Babajide PUBLISH A7AUTY@@cfp.bsidesvienna.at -A7AUTY Hacking Mobile Apps in a Structured Way en en 20260627T110000 20260627T114500 0.04500 Hacking Mobile Apps in a Structured Way PUBLIC CONFIRMED Medium Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/A7AUTY/ Kreativraum 3.1 (Track 3 - Women4Cyber/Rookie) Sven Schleier PUBLISH 8ACFFK@@cfp.bsidesvienna.at -8ACFFK HackerOne Club Austria en en 20260627T130000 20260627T133000 0.03000 HackerOne Club Austria HackerOne Club Austria! This club is dedicated to building a collaborative and welcoming hacker community across Austria. Whether you are an experienced bug bounty hunter, cybersecurity professional, student, or someone just getting started in ethical hacking, you are welcome to join and grow with the community. The Austria HackerOne Club organizes a wide range of events including: - local meetups - live hacking sessions - educational workshops - Capture The Flag (CTF) competitions - conference gatherings - world cup meetings (AWC) - discussions focused on bug bounty, ethical hacking, and cybersecurity. These events are designed to encourage hands-on learning, knowledge sharing, collaboration, and stronger connections within the local security community. Hacking Meeting: The hacking meetings are especially interesting to community members. Hackers come together in person or virtually to hack on bug bounty programs. HackerOne may partner with private programs to provide an interesting meetup. Members may get access to new and fresh hacking scope or invitations to private programs. HackerOne provides a leaderboard to track the progress of the meetup. There is the possibility of special reward multipliers or better payouts for researchers. By joining this club, you'll stay informed about upcoming HackerOne community events in Austria, connect with fellow hackers in your region, and become part of a global community that values learning, responsible disclosure, and collaboration. PUBLIC CONFIRMED Short Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/8ACFFK/ Kreativraum 3.1 (Track 3 - Women4Cyber/Rookie) Alexander Krenn PUBLISH VA3RKU@@cfp.bsidesvienna.at -VA3RKU Offensive AI: Red Teaming Machine Learning Systems en en 20260627T133500 20260627T142000 0.04500 Offensive AI: Red Teaming Machine Learning Systems This talk introduces the Offensive AI Framework, an open, structured knowledge base mapped to OWASP ML Top 10, OWASP LLM Top 10, and Google's SAIF. There will be a public repository where all the information related to the frameworks will be matched between the frameworks. Attendees will learn how ML models are trained and where that process introduces risk, how the major AI vulnerability frameworks are organized, and see practical exploitation of real vulnerabilities including data poisoning, prompt injection, and model fingerprinting. The session closes with a fully local live demo, no internet required, showing how reconnaissance leads to successful exploitation through indirect prompt injection (I will have a recorded video of the exploitation just in case as well). [Offensive-AI Framework](https://github.com/hardsoftsecurity/Offensive-AI) PUBLIC CONFIRMED Medium Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/VA3RKU/ Kreativraum 3.1 (Track 3 - Women4Cyber/Rookie) David De Maya Merras PUBLISH XX8GYX@@cfp.bsidesvienna.at -XX8GYX XPC Client Validation? Music to my ears! en en 20260627T150500 20260627T160500 1.00000 XPC Client Validation? Music to my ears! PUBLIC CONFIRMED Medium Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/XX8GYX/ Kreativraum 3.1 (Track 3 - Women4Cyber/Rookie) Florian Haselsteiner PUBLISH 78ZAJJ@@cfp.bsidesvienna.at -78ZAJJ What is software? en en 20260627T161000 20260627T171000 1.00000 What is software? PUBLIC CONFIRMED Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/78ZAJJ/ Kreativraum 3.1 (Track 3 - Women4Cyber/Rookie) Mária Kamilla Huszár PUBLISH V7JZEW@@cfp.bsidesvienna.at -V7JZEW Parsing CBOR is a Minefield: A Study of CBOR Parser behavior en en 20260627T171500 20260627T174500 0.03000 Parsing CBOR is a Minefield: A Study of CBOR Parser behavior CBOR underpins much of modern security infrastructure. Yet whether two implementations agree on what a given byte sequence means, and how they handle edge cases and malformed data, has received little public systematic attention. This talk examines 11 widely-used CBOR parsers spanning C, Rust, Python, Java, Go, C#, and JavaScript, selected for real-world deployment, active maintenance, and architectural diversity. I cover the testing methodology, including fuzzing and a differential-testing harness that runs shared valid and invalid inputs across all parsers, as well as how the resulting divergences are visualized and analyzed. Finally, I discuss the observed divergences and any crashes the testing uncovered, their underlying reasons, and the security implications. As this talk presents ongoing research as part of my master's thesis, the results are still preliminary and open to change. PUBLIC CONFIRMED Short Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/V7JZEW/ Kreativraum 3.1 (Track 3 - Women4Cyber/Rookie) Jakob Pachmann PUBLISH FP9L8D@@cfp.bsidesvienna.at -FP9L8D NTLM-Relaying in Practice en en 20260627T093000 20260627T113000 2.00000 NTLM-Relaying in Practice In a Windows Active Directory environment, minor misconfigurations can lead to total domain compromise. This hands-on workshop dives into NTLM-Relaying attacks and explores exactly what happens when critical defenses like SMB signing and LDAP signing are left disabled and how NTLM-Relaying can still be abused with these mechanisms enabled. ​Through practical, real-world demonstrations, participants will learn: <ul> <li>How attackers intercept and abuse NTLM authentication</li> <li>Lateral movement and privilege escalation scenarios caused by missing mitigations</li> <li>Step-by-step blueprints to effectively secure your AD environment against these threats</li> </ul> ​By the end of this session, you will not only understand how attackers exploit NTLM-Relaying but also how to shut them down. To attend this workshop, participants will need a laptop with an installed SSH client. PUBLIC CONFIRMED Workshop https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/FP9L8D/ Kleiner Saal (Workshops Track) Benjamin Floriani Patrick Pongratz PUBLISH 8Q9BGK@@cfp.bsidesvienna.at -8Q9BGK Try to open a lock keyless en en 20260627T123000 20260627T142900 1.05900 Try to open a lock keyless PUBLIC CONFIRMED Workshop https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/8Q9BGK/ Kleiner Saal (Workshops Track) deac TiborElias PUBLISH 9JKBGC@@cfp.bsidesvienna.at -9JKBGC Before There Was a Password — The Philosophy and Politics of Secrecy en en 20260627T151500 20260627T160000 0.04500 Before There Was a Password — The Philosophy and Politics of Secrecy PUBLIC CONFIRMED Medium Talk https://cfp.bsidesvienna.at/bsidesvienna-0x7ea/talk/9JKBGC/ Kleiner Saal (Workshops Track) Iryna