Talks

Full kill-chain dissection of telecom malware across RAN, core network, signaling plane, and supply chain attack surfaces. Covered network-resident implants in MME/AMF/UPF, SS7/Diameter/GTP signaling abuse, rogue eNodeB/gNodeB firmware, and SIM/USIM backdoors. Case studies include LightBasin (multi-MNO intrusion across 13+ operators) and a modified PGW binary implant surviving firmware upgrades. Detection pipelines using canary IMSIs, anomaly-based signaling telemetry, and future threats spanning ORAN xApp supply chain and AI-driven social engineering.

Technical walkthrough of CVE-2024-9680 analysis methodology — from fuzzing-induced discovery to a weaponized, reliable local privilege escalation exploit targeting Linux 6.x kernels.

Full attack chain against modern ATM disk encryption — bypassing cold-boot protections, extracting keys from volatile memory, and defeating FDE implementations in production cash-handling systems. Covered firmware extraction techniques, unattended key management flaws, and the physical attack surface that survives software hardening. Concluded with responsible disclosure findings and recommendations for hardened boot architectures.

Security analysis of networked vending machines running embedded Linux and Windows IoT — from unauthenticated cashless payment APIs to remote code execution via unpatched telemetry agents. Demonstrated full compromise of MDB (Multi-Drop Bus) payment interfaces, manipulation of product pricing logic, and persistent implants surviving reboot cycles. Findings cover 5+ major vending platforms and responsible disclosure outcomes with vendors.

Systematic analysis of JWT library implementations across 40+ major frameworks, revealing authentication bypass vulnerabilities in production SSO deployments.

Presentation of the IoT exploitation framework — architecture, MQTT fuzzing methodology, and the discovery pipeline that surfaced 12 critical CVEs in consumer smart home devices.

Methodology behind discovering 25+ LLM vulnerabilities with an 80% guardrail bypass rate — systematic adversarial prompting, model behaviour analysis, and responsible disclosure to AI vendors.

Top-ranked nationally and 1st at university level. Competed across algorithms, systems, and security challenges under a 24-hour competitive format.