BitterSecurity/Decepticon - Page 4

data-and-model-poisoning

Hunt LLM training-data and model poisoning (OWASP LLM04:2025) — adversarial inputs that bias future model behaviour through fine-tuning, RLHF, or continuous-learning loops.

Langue du texte source : anglais

deserialization

Hunt insecure deserialization (CWE-502) across Python pickle, Java ObjectInputStream / Jackson / SnakeYAML, .NET BinaryFormatter / DataContractJson, PHP unserialize, Ruby Marshal/YAML.load, and Node.js vm. Direct path to unauthenticated RCE.

Langue du texte source : anglais

excessive-agency

Hunt LLM excessive agency (OWASP LLM06:2025) — agentic systems granted too many tools, too broad permissions per tool, or unsupervised authority to act on the user / business behalf, producing financial loss, data loss, or destructive operations from a single…

Langue du texte source : anglais

idor

Hunt Insecure Direct Object Reference (CWE-639) — missing authorization checks on object IDs. Covers horizontal vs vertical privilege escalation, UUID vs integer guessing, and GraphQL introspection-driven IDOR discovery.

Langue du texte source : anglais

improper-output-handling

Hunt improper LLM output handling (OWASP LLM05:2025) — downstream code that trusts unstructured model output and renders / executes / shells it without sanitisation, producing XSS, SSRF, SQL injection, RCE, and SSTI via the model channel.

Langue du texte source : anglais

misinformation

Hunt LLM misinformation / overreliance (OWASP LLM09:2025) — confident-but-wrong outputs that flow into downstream automated decisions, compliance reports, customer communications, or autonomous code commits without verification.

Langue du texte source : anglais

path-traversal

Hunt directory traversal and archive traversal (ZipSlip/TarSlip) from user input to filesystem operations.

Langue du texte source : anglais

prompt-injection

Hunt LLM prompt injection and tool-call hijacking in modern AI-integrated applications (CWE-1427). Covers indirect injection via RAG, tool abuse, exfiltration chains, and jailbreak-to-RCE pivots on agentic systems.

Langue du texte source : anglais

prototype-pollution

Hunt JavaScript prototype pollution (CWE-1321) — the 2023-2026 meta-vulnerability that chains into RCE, auth bypass, and SSRF on most Node.js stacks.

Langue du texte source : anglais

sensitive-information-disclosure

Hunt LLM sensitive-information disclosure (OWASP LLM02:2025) — leakage of PII, secrets, internal source, model details, and other-tenant data through model outputs, training-data extraction, or retrieval-side joins.

Langue du texte source : anglais

sql-injection

Hunt SQL injection (CWE-89) via source-level taint tracking. Covers string concat, format-string, ORM raw queries, second-order injection, and NoSQL injection in MongoDB/DynamoDB.

Langue du texte source : anglais

ssrf

Hunt Server-Side Request Forgery (CWE-918) through taint analysis from user-controlled URLs to HTTP client sinks. Covers cloud metadata pivoting, DNS rebinding, gopher smuggling, and the IMDSv1 → IAM role chain that turns SSRF into RCE.

Langue du texte source : anglais

ssti

Hunt server-side template injection across Jinja2/Twig/Freemarker/Velocity/Handlebars and validate progression from expression injection to code execution.

Langue du texte source : anglais

system-prompt-leakage

Hunt LLM system-prompt leakage (OWASP LLM07:2025) — exfiltration of the privileged system prompt revealing internal rules, secrets baked in, tool inventory, and business logic that should not be client-visible.

Langue du texte source : anglais

trust-boundary-analysis

Trust boundary mapping and startup sequence audit for developer tools, CLI apps, and plugin systems. Load when the target is a developer tool, CLI, IDE extension, or any application that loads config from the current directory.

Langue du texte source : anglais

unbounded-consumption

Hunt LLM unbounded consumption (OWASP LLM10:2025) — denial-of-wallet and denial-of-service against LLM endpoints via unrestricted prompt size, runaway tool loops, expensive model selection, and unauthenticated fan-out.

Langue du texte source : anglais

vector-and-embedding-weaknesses

Hunt vector / embedding weaknesses (OWASP LLM08:2025) — adversarial inputs against the RAG / similarity layer that cause cross-tenant leak, embedding-inversion privacy loss, semantic confusion, and retriever-driven prompt injection.

Langue du texte source : anglais

xxe

Hunt XML External Entity flaws in parsers and validate file read / SSRF impact with strict negative controls.

Langue du texte source : anglais

references

External knowledge integration — HackerOne reports, PayloadsAllTheThings, Book of Secret Knowledge, CVE PoC corpora, bug bounty methodologies, and reference pentest agent architectures. Use these to calibrate, look up payloads, and accelerate research.

Langue du texte source : anglais

kill-chain-analysis

Kill chain analysis and attack path decision-making — findings analysis, attack vector selection, target prioritization, phase transitions.

Langue du texte source : anglais

detector-overview

Stage 2 vulnerability detector playbook. Reads source around CANDIDATE nodes and promotes real bugs to VULNERABILITY + HYPOTHESIS. Read-only. Load at detector-agent startup.

Langue du texte source : anglais

exploiter-overview

Stage 5 exploit construction playbook. Weaponizes validated primitives into multi-step chains that reach crown jewels. Load at exploiter-agent startup.

Langue du texte source : anglais

patcher-overview

Stage 4 patch generation playbook. Minimal diffs for validated findings with mandatory patch_verify. Load at patcher-agent startup.

Langue du texte source : anglais

scanner-overview

Stage 1 broad-spectrum scanner playbook. Sharded sweep over very large codebases producing CANDIDATE nodes for the Detector to reason about. Load at scanner-agent startup.

Langue du texte source : anglais

bounty-report-formatter

Bug bounty report formatting for HackerOne, Bugcrowd, Immunefi, and GitHub Security Advisories. Load after validate_finding succeeds and the finding needs to be submitted to a bounty program.

Langue du texte source : anglais

vulnresearch-orchestrator

Five-stage modular vulnerability pipeline orchestrator. Delegates scan → detect → verify → patch → exploit through OPPLAN objectives. Load at orchestrator startup.

Langue du texte source : anglais

benchmark

Benchmark mode marker — engagement objective is flag capture. Generic engagement rules apply unchanged.

Langue du texte source : anglais

supply-chain

Hunt LLM supply-chain compromise (OWASP LLM03:2025) — malicious or backdoored models, datasets, adapters, plugins, MCP servers, and tokenizer / framework dependencies that ship inside an AI-integrated product.

Langue du texte source : anglais

iot-overview

Use when the engagement target is IoT, embedded Linux, RTOS, or any device reachable via UART/JTAG/SWD or by extracting its firmware. Covers firmware acquisition, binwalk extraction, filesystem mounting, default-credential hunting, bootloader attacks,…

Langue du texte source : anglais

pentest-task-tree

Iterative PTT (Penetration Testing Tree) session reasoning — build, update, and traverse a live numbered task tree to drive LLM-guided pentest decisions across a full session.

Langue du texte source : anglais

access-control

Missing modifiers, wrong msg.sender checks, default-public functions, missing onlyOwner / onlyRole / onlyDAO authorization.

Langue du texte source : anglais

flash-loan

Flash-loan exploit patterns — callback reentrancy, oracle amplification, governance attacks, unauthenticated callback handlers.

Langue du texte source : anglais

oracle-manipulation

Hunt single-block oracle manipulation — spot-price AMM oracles, manipulable TWAP, dependent calculations, missing staleness checks.

Langue du texte source : anglais

reentrancy

Hunt and validate reentrancy bugs — classic cross-function, same-function, read-only, and cross-contract variants.

Langue du texte source : anglais

signature-replay

Signature replay attacks — missing nonces, missing chain ID, ecrecover zero address, signature malleability, cross-chain replay.

Langue du texte source : anglais

upgradeable-proxy

Proxy upgrade patterns and their bugs — uninitialized implementation, storage slot collisions, selector clashes, unprotected upgrade auth.

Langue du texte source : anglais

aws-iam-enum

Enumerate AWS IAM policies, detect privilege escalation paths per Rhino Security Labs canonical 21 primitives.

Langue du texte source : anglais

imds-pivot

Pivot from SSRF or RCE to cloud Instance Metadata Service (IMDS) — extract IAM role creds, instance identity, user-data secrets.

Langue du texte source : anglais

k8s-pivot

Kubernetes attack playbook — service-account token theft, RBAC abuse, pod escape, hostPath mount abuse, kube-api-server pivoting.

Langue du texte source : anglais

s3-takeover

Detect and claim dangling S3 buckets referenced by subdomains (CNAME → s3 hostnames where bucket no longer exists).

Langue du texte source : anglais