Security Tester

Department: AIDA

Experience: 2-5

• • Own end-to-end VAPT engagements across web apps, REST/GraphQL APIs, cloud infrastructure, and internal networks — from scoping and threat modelling to exploitation, reporting, and retesting.
  • Red-team AI-native applications: prompt injection (direct and indirect), jailbreaks, system-prompt leakage, insecure output handling, RAG poisoning, training-data extraction, model denial-of-service, and excessive-agency abuse in agentic systems.
  • Break agents and their tool-chains: test tool/function-calling boundaries, MCP (Model Context Protocol) server exposure, sandbox escapes, and privilege escalation through chained tool calls.
  • Map findings to real frameworks — OWASP Top 10, OWASP LLM Top 10, OWASP Agentic Security (ASI), MITRE ATT&CK, and MITRE ATLAS — so remediation is grounded, not hand-wavy.
  • Score and prioritize vulnerabilities using CVSS, with clear, reproducible proof-of-concept and business-impact context that both engineers and leadership can act on.
  • Write reports people read: crisp, developer-friendly write-ups with reproduction steps, evidence, and pragmatic remediation guidance — not a wall of scanner output.
  • Shift security left: help embed automated security testing (SAST, SCA, DAST, secret scanning, IaC scanning, LLM red-team checks) into CI/CD pipelines and PR gates.
  • Retest and verify fixes, track remediation to closure, and partner with engineering to make the fix stick.
  • Stay ahead of the curve — track emerging AI attack techniques and feed new test cases back into our internal red-team playbooks.

• • 2.5+ years of hands-on offensive security / penetration testing experience.
  • Demonstrated experience with the full VAPT lifecycle for web, API, and cloud targets — you can scope, exploit, and communicate, not just run a tool.
  • Strong grasp of the OWASP Top 10 and common web/API vulnerability classes (authn/authz flaws, injection, SSRF, IDOR, deserialization, misconfigurations).
  • Working knowledge of LLM and AI application attack surfaces — prompt injection, jailbreaks, insecure output handling, and the OWASP LLM Top 10 — or a clear, provable appetite to go deep here fast.
  • Hands-on with core offensive tooling: Burp Suite, OWASP ZAP, Nmap, Nuclei, Metasploit, sqlmap, and similar.
  • Comfortable in at least one scripting language (Python strongly preferred) to build custom exploits, harnesses, and automation.
  • Solid understanding of web protocols, authentication/authorization (OAuth 2.0 / OIDC, JWT, session management) and how they break.
  • Familiarity with cloud security fundamentals (AWS, Azure, or GCP) and containerized environments (Docker, Kubernetes).
  • Ability to write clear, prioritized, reproducible reports and explain risk to both engineers and non-technical stakeholders.

• • Hands-on experience with LLM/agent red-teaming frameworks: Garak, PyRIT, DeepTeam, promptfoo, or Microsoft’s AI Red Teaming Agent.
  • Experience with AI guardrail / defense tooling: Lakera Guard, Rebuff, Llama Prompt Guard, or custom prompt-injection classifiers.
  • Familiarity with MCP security and agent-sandboxing concepts (MCP Scan, tool-permission boundaries, isolation).
  • Exposure to RAG pipeline security, vector-store poisoning, and multi-agent orchestration risks.
  • Knowledge of MITRE ATLAS and the NIST AI Risk Management Framework (AI RMF).
  • Experience integrating security testing into CI/CD (GitHub Actions, GitLab CI) and policy-as-code gating (OPA / Rego).
  • Cloud-native security experience with Kubernetes (RBAC, network policies, admission control).
  • Bug-bounty track record, CTF wins, published research, or CVEs to your name.