Microsoft Security Response Center

Thank you to everyone who joined us this week for the second edition of BlueHat India. BlueHat is more than just a conference, it’s a community. One where the security community from inside and outside Microsoft come together as peers to share, challenge, and learn from one another. From deep technical talks to hallway debates, this year’s BlueHat India showed how collaboration drives progress in security. Together, we're helping build a safer, more secure world for everyone.   Special shoutout to the BlueHat organizers and volunteers. This event would not be possible without your energy, dedication, and behind-the-scenes magic. Until next year. 💙 #BlueHatIndia

We kicked off Day 2 of BlueHat India with opening remarks from Charu Srinivasan, CVP of Engineering at Microsoft, who explored the rise of Agentic AI: autonomous agents that will soon play a role in nearly every system we build. Her message was clear: defenders must secure these agents and use them to strengthen security itself.   Ram Shankar Siva Kumar 🦝, Data Cowboy & Head of Microsoft’s AI Red Team, delivered a powerful keynote unpacking lessons from red teaming over 100 generative AI systems. From model manipulation to psycho-social harms, Ram showed how attackers are evolving and why security must evolve with them.   The rest of the day featured powerful sessions from across the community:   • Shriya Maniktala introduced Agentic AI red teaming, where LLM-powered adversaries adapt in real-time using live feedback and tools like AETHR • Manish Gupta & Yash Bharadwaj showed how adversaries are using serverless platforms to build stealthy phishing kits that slip past traditional defenses • Omkar Joshi & Pallavi Deshmukh revealed just how easily LLM safety guardrails can be bypassed, calling for adversarial testing and layered defenses • Rituraj Jodha discussed how graph-based detection and Graph Neural Networks can surface malicious OAuth apps that hide in plain sight • Sagar Bhure showed how deepfake detection systems can be defeated with subtle tweaks and adversarial image prompts • Ashish Dhone shared lessons from earning over $50K in blind XSS bounties, highlighting overlooked surfaces like feedback forms and custom dashboards   To every attendee, speaker, and partner: thank you. Your dedication to learning together, sharing openly, and supporting one another is what makes BlueHat special. #BlueHatIndia

At BlueHat India Day 2, Ram Shankar Siva Kumar 🦝, Microsoft’s resident Data Cowboy and Head of the AI Red Team, took the stage for a keynote that was equal parts eye-opening and unsettling. Ram unpacked what happens when adversarial machine learning meets generative AI. From subtle image perturbations that fool models into seeing gibbons instead of pandas, to video analytics systems that could be manipulated through outdated dependencies, the message was clear: AI systems remain deeply vulnerable.   He shared lessons from red teaming over 100 generative AI systems, outlining the four harm categories that Microsoft’s AI Red Team probes for: • Traditional security risks • Content safety issues • Dangerous capabilities   Five takeaways that stuck with us: 1. Long before AI Safety craze, researchers have been warning about how ML systems are not built with security in mind 2. Traditional security failures do not go away 3. Attackers don’t always use gradients 4. Psycho-social harms are a new frontier 5. Agents are also going to be messy   Ram reminded us that while generative AI brings powerful new tools, it also expands the attack surface. Security must evolve to meet this challenge, and that starts with understanding how these systems can be broken.   #BlueHatIndia

At BlueHat India Day 2, Charu Srinivasan, CVP of Engineering at Microsoft, opened the morning with a powerful message about one of the biggest shifts reshaping our field: Agentic AI.   In this new era, intelligent agents won’t just assist, they’ll become ubiquitous across our systems, making decisions, interacting with data, and collaborating with other agents and humans alike. As Charu noted, the more capable these agents become, the more critical it is that security evolves to keep pace, operating at AI speed. She outlined a future where AI agents proactively identify data risks, apply dynamic access controls, and support security by design. But this transformation brings complexity, security professionals must now consider identity, permissions, privacy, insider risk, threat protection, and governance in a whole new light. Charu emphasized a dual mission for our community: 1.  AI agents for security 2. Security for AI agents This relationship is deeply interconnected and defining it will be one of our biggest challenges in the coming years. As autonomous agents grow more powerful, defenders must not only secure them, but guide and govern their actions. Agentic AI will reinvent our roles, but human traits strategic thinking, curiosity, empathy, and diverse perspectives will remain irreplaceable. Charu left us with a call to action: Developing your skills as an AI leader isn’t optional, it’s how we’ll shape the future of security. #BlueHatIndia

Day 1 of BlueHat India brought together a passionate, global community of security professionals, from seasoned defenders to up-and-coming researchers, who came to learn, share, and connect. We kicked off with keynotes from Tom Gallagher and David Weston who discussed everything from Zero Day Quest to the risks of on-device AI, followed by our sessions: • George Hughey discussed how Microsoft finds variant vulnerabilities using competition research • Ali Abdollahi walked through red teaming mobile networks with lessons from real-world telecom breaches • Vaibhav Deshmukh & Sudhir Devkar revealed the inner workings of ransomware group Black Basta through leaked operational chats • Sathwik Ram Prakki discussed APT SideCopy’s evolution and targeting shifts • Joshua Bahirvani & Shaleen Dev P.K. showed how attackers abuse dormant cloud identities in stealthy campaigns • Anurag Bohra & Anant Pulgam shared new techniques used by the XCSSET macOS malware • Thirumalai Natarajan.M discussed threats against virtual infrastructure and hypervisors   Beyond the main stage, attendees explored hands-on learning across six Villages: • AI in Security Village: Attendees tested their skills in AI-themed CTFs and explored how AI is reshaping the security landscape • Garage Village: From soldering workshops to live 3D printing, curious makers got hands-on with hardware prototyping • Forensics Village: Participants dove into digital forensics challenges, quizzes, and capture-the-flag scenarios • MSRC Village: Bug hunting, insecure code spotting, and deep dives into Microsoft’s bounty programs kept researchers engaged • AppSec Village: Secure coding exercises, quizzes, and reverse engineering challenges gave developers a chance to sharpen their skills • Arena: Thought-provoking fireside chats sparked discussion on Diversity & Inclusion and what quantum computing means for the future of security To everyone who presented, asked questions, sparked hallway conversations, and shared on social—thank you. This is what the BlueHat community is all about. Day 2 is up next in just a few short hours. #BlueHatIndia

Think AI safety is solved with guardrails? Not quite. Omkar Joshi and Pallavi Deshmukh from Coupa Software will show how Large Language Models are still getting tricked, and the clever ways attackers are bypassing protections. Join them tomorrow at BlueHat India to learn about real-world hacks and what we can do to outsmart them. #BlueHatIndia

The BlueHat India Villages are now open! Swing by between sessions and get hands-on, challenge your skills, and connect with security experts. 𝗔𝗜 𝗶𝗻 𝗦𝗲𝗰𝘂𝗿𝗶𝘁𝘆 𝗩𝗶𝗹𝗹𝗮𝗴𝗲: Explore AI’s role in cybersecurity and test yourself with AI-themed CTF challenges 𝗚𝗮𝗿𝗮𝗴𝗲 𝗩𝗶𝗹𝗹𝗮𝗴𝗲: Try soldering workshops and live 3D printing demos, perfect for prototyping newbies 𝗙𝗼𝗿𝗲𝗻𝘀𝗶𝗰𝘀 𝗩𝗶𝗹𝗹𝗮𝗴𝗲: Learn digital forensics techniques, take a quiz, and join the CTF 𝗠𝗦𝗥𝗖 𝗩𝗶𝗹𝗹𝗮𝗴𝗲: Hunt bugs, spot insecure code, dive into a phishing CTF, and get the scoop on bounty programs 𝗔𝗽𝗽𝗦𝗲𝗰 𝗩𝗶𝗹𝗹𝗮𝗴𝗲: Brush up on secure coding, take on quizzes, and test yourself in CTF + RTF challenges 𝗔𝗿𝗲𝗻𝗮: Don’t miss fireside chats on Diversity & Inclusion and the future of Quantum Computing #BlueHatIndia

David Weston, CVP, Security at Microsoft, kicked off Day 1 of BlueHat India with a keynote on the security implications of on-device AI and the evolving threat landscape. He outlined three critical categories of AI risk: 1. Data attacks: The biggest threat isn’t AI itself, but the sensitive data it needs to function. As devices store more data for AI processing, they become high-value exfiltration targets. 2. Model attacks: AI is increasingly in the decision-making path. Real-world examples like Windows Hello show how tampering with models can lead to direct security breaches. 3. Abuse: The fastest-growing category, where attackers misuse AI capabilities for unintended purposes. David also shared insights into Microsoft’s latest features like Windows Recall, designed to boost productivity with contextual AI. But with that power comes risk. He emphasized the need for innovations like biometric proof-of-presence, encrypted data isolation, tamper-proof settings, and confidential computing to defend against increasingly sophisticated threats, including those from users with legitimate access. #BlueHatIndia

Tom Gallagher, VP of Engineering at MSRC, opened BlueHat India with a question: “Anyone here a threat actor?” He went on to highlight the recent Microsoft Zero Day Quest, which saw over 100 researchers qualify. Today, we're proud to welcome three standout researchers from India—Niraj Mahajan, Ashish Dhone, and Anto Denvo. And we’re just getting started. Submit your vulnerabilities to MSRC and join us in protecting customers, Microsoft, and the global community. #BlueHatIndia

Good morning, BlueHat India! Registration is now open, and we’re thrilled to welcome you to the second edition of the conference. Who’s excited for Day 1? Head over to the registration desk, grab some breakfast, and get ready for opening remarks from Tom Gallagher, VP of Engineering at MSRC at 10:05am, followed by the Day 1 keynote from David Weston. #BlueHatIndia