{ "info": { "_postman_id": "ptai-api-collection", "name": "ptAI API", "description": "Curated requests for every ptAI attack, mirroring the dashboard's **Data type → Family → Scenario → Algorithm** hierarchy.\n\n**Before sending:** open the *Body* tab, scroll to the file slots (`dataset`, `model`, `scaler`, `encoder` for structured; `model` and optional `dataset_zip` for image), and click *Select Files* to attach your assets. `params` is pre-filled with the dashboard's default values for the chosen algorithm + data type.\n\n**Response:** every attack returns a **ZIP** by default (all results + artifacts). In Postman, use *Save Response → Save to a file* to keep it. Add `?format=json` to the URL if you'd rather get the JSON envelope inline.\n\n**Slow attacks:** every attack runs as a background job. Fast ones still return inline like above; slow ones (jsma, hsj, cw, zoo, boundary, pentest, inference...) return `202 {job_id, ...}` instead — see the **Jobs (async attacks)** folder to poll/fetch/cancel.\n\n**Auth:** set the `apiKey` collection variable (Variables tab) to your `X-API-Key` — every request in this collection sends it automatically via the collection-level Authorization.", "schema": "https://schema.getpostman.com/json/collection/v2.1.0/collection.json" }, "item": [ { "name": "Meta", "description": "Health, schema, and documentation endpoints. Most are always public — no API key required — except where noted.", "item": [ { "name": "Health check", "request": { "method": "GET", "header": [], "url": { "raw": "{{baseUrl}}/healthz", "host": [ "{{baseUrl}}" ], "path": [ "healthz" ] } } }, { "name": "OpenAPI schema", "request": { "method": "GET", "header": [], "url": { "raw": "{{baseUrl}}/openapi.json", "host": [ "{{baseUrl}}" ], "path": [ "openapi.json" ] } } }, { "name": "Landing page", "request": { "method": "GET", "header": [], "url": { "raw": "{{baseUrl}}/", "host": [ "{{baseUrl}}" ], "path": [] } } }, { "name": "User API Guide (Markdown download)", "request": { "method": "GET", "header": [], "url": { "raw": "{{baseUrl}}/readme", "host": [ "{{baseUrl}}" ], "path": [ "readme" ] } } }, { "name": "Validate API key (whoami)", "request": { "method": "GET", "header": [], "url": { "raw": "{{baseUrl}}/api/auth/whoami", "host": [ "{{baseUrl}}" ], "path": [ "api", "auth", "whoami" ] }, "description": "Unlike the other requests in this folder, this one requires a valid `X-API-Key` (set the `apiKey` collection variable first) — use it to check a pasted key works before running an attack." } } ] }, { "name": "Unified (one full-config file)", "description": "`POST /api/attack` — one JSON config describes the whole request (data_type + family + scenario/attack + params). Put it inline in `config`, or upload a `.json` in `config_file` (download ready-made ones from /templates). Then attach your data files. Same ZIP/JSON response as the granular endpoints.", "item": [ { "name": "Evasion PGD (structured) — inline config", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "config", "value": "{\"data_type\":\"structured\",\"family\":\"evasion\",\"scenario\":\"white\",\"attack\":\"pgd\",\"columns_to_drop\":\"flow_id,src_ip,dst_ip,src_port,dst_port\",\"label_col\":\"label\",\"params\":{\"eps\":0.1,\"eps_step\":0.01,\"max_iter\":40}}", "type": "text" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } }, { "name": "Evasion FGSM (image) — inline config", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "config", "value": "{\"data_type\":\"image\",\"family\":\"evasion\",\"scenario\":\"white\",\"attack\":\"fgsm\",\"source\":\"mnist\",\"limit\":200,\"params\":{\"eps\":0.03}}", "type": "text" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" } ] } } }, { "name": "Poisoning Backdoor (structured) — inline config", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "config", "value": "{\"data_type\":\"structured\",\"family\":\"poisoning\",\"attack\":\"backdoor\",\"columns_to_drop\":\"flow_id,src_ip,dst_ip,src_port,dst_port\",\"label_col\":\"label\",\"params\":{\"target_class\":0,\"trigger_features\":\"-1\",\"trigger_value\":1.0,\"poison_fraction\":0.1}}", "type": "text" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } }, { "name": "Any attack — upload config_file instead", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack" ] }, "description": "Leave `config` empty and upload a full-config .json from /templates in `config_file`.", "body": { "mode": "formdata", "formdata": [ { "key": "config", "value": "{}", "type": "text" }, { "key": "config_file", "type": "file", "src": [], "description": "Full-config .json from /templates (config set)" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } } ] }, { "name": "Structured (tabular)", "description": "Tabular attacks — CSV + Keras model + scaler + encoder.", "item": [ { "name": "Evasion", "description": "Adversarial inputs that fool the model at inference time.", "item": [ { "name": "White-box (gradient access)", "item": [ { "name": "FGSM", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/whitebox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "whitebox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "attack", "value": "fgsm", "type": "text" }, { "key": "params", "value": "{\"eps\":0.1}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } }, { "name": "PGD", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/whitebox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "whitebox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "attack", "value": "pgd", "type": "text" }, { "key": "params", "value": "{\"eps\":0.1,\"eps_step\":0.01,\"max_iter\":40}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } }, { "name": "BIM", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/whitebox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "whitebox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "attack", "value": "bim", "type": "text" }, { "key": "params", "value": "{\"eps\":0.1,\"eps_step\":0.01,\"max_iter\":10}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } }, { "name": "JSMA", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/whitebox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "whitebox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "attack", "value": "jsma", "type": "text" }, { "key": "params", "value": "{\"theta\":1.0,\"gamma\":0.1}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } }, { "name": "CW", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/whitebox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "whitebox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "attack", "value": "cw", "type": "text" }, { "key": "params", "value": "{\"confidence\":0.0,\"learning_rate\":0.001,\"binary_search_steps\":10,\"max_iter\":20,\"initial_const\":0.9,\"batch_size\":32}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } } ] }, { "name": "Black-box (decision-only)", "item": [ { "name": "ZOO", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/blackbox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "blackbox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "attack", "value": "zoo", "type": "text" }, { "key": "params", "value": "{\"confidence\":0.5,\"learning_rate\":0.01,\"max_iter\":20,\"nb_parallel\":20}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } }, { "name": "Boundary", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/blackbox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "blackbox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "attack", "value": "boundary", "type": "text" }, { "key": "params", "value": "{\"max_iter\":1000,\"epsilon\":0.01}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } }, { "name": "HopSkipJump", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/blackbox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "blackbox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "attack", "value": "hsj", "type": "text" }, { "key": "params", "value": "{\"max_iter\":25,\"max_eval\":10000}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } } ] }, { "name": "Grey-box (gradients on surrogate)", "item": [ { "name": "FGSM", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/greybox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "greybox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "attack", "value": "fgsm", "type": "text" }, { "key": "params", "value": "{\"eps\":0.1}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } }, { "name": "PGD", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/greybox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "greybox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "attack", "value": "pgd", "type": "text" }, { "key": "params", "value": "{\"eps\":0.1,\"eps_step\":0.01,\"max_iter\":40}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } }, { "name": "BIM", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/greybox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "greybox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "attack", "value": "bim", "type": "text" }, { "key": "params", "value": "{\"eps\":0.1,\"eps_step\":0.01,\"max_iter\":10}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } }, { "name": "JSMA", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/greybox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "greybox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "attack", "value": "jsma", "type": "text" }, { "key": "params", "value": "{\"theta\":1.0,\"gamma\":0.1}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } }, { "name": "CW", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/greybox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "greybox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "attack", "value": "cw", "type": "text" }, { "key": "params", "value": "{\"confidence\":0.0,\"learning_rate\":0.001,\"binary_search_steps\":10,\"max_iter\":20,\"initial_const\":0.9,\"batch_size\":32}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } } ] } ] }, { "name": "Inference (Membership)", "description": "Test whether the model leaks its training-set membership.", "item": [ { "name": "MIA · Random Forest meta-classifier", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/inference", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "inference" ] }, "description": "Rule-Based + Black-Box MIA run in parallel. Change `attack_model_type` to gb or lr to swap the meta-classifier.", "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "attack_model_type", "value": "rf", "type": "text", "description": "rf · gb · lr" }, { "key": "params", "value": "{\"split_ratio\":0.7,\"max_samples\":1000,\"seed\":42}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } } ] }, { "name": "Extraction (CopycatCNN)", "description": "Train a stolen ('thieved') copy of the model from victim queries.", "item": [ { "name": "CopycatCNN", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/extraction", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "extraction" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "params", "value": "{\"nb_stolen\":500,\"nb_epochs\":5,\"batch_size\":64,\"use_probability\":true}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } } ] }, { "name": "Poisoning", "description": "Inject malicious training samples, retrain a clone, measure ASR.", "item": [ { "name": "Backdoor", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/poisoning", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "poisoning" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "attack", "value": "backdoor", "type": "text" }, { "key": "params", "value": "{\"target_class\":0,\"trigger_features\":\"-1\",\"trigger_value\":1.0,\"poison_fraction\":0.10,\"nb_epochs\":8,\"batch_size\":64,\"seed\":42}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } }, { "name": "Clean-Label", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/poisoning", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "poisoning" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "attack", "value": "clean_label", "type": "text" }, { "key": "params", "value": "{\"target_class\":0,\"trigger_features\":\"-1\",\"trigger_value\":1.0,\"poison_fraction\":0.10,\"pp_poison\":0.33,\"eps\":0.3,\"eps_step\":0.1,\"max_iter\":100,\"nb_epochs\":8,\"batch_size\":64,\"seed\":42}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } }, { "name": "Hidden Trigger", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/poisoning", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "poisoning" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text" }, { "key": "attack", "value": "hidden_trigger", "type": "text" }, { "key": "params", "value": "{\"target_class\":0,\"source_class\":1,\"trigger_features\":\"-1\",\"trigger_value\":1.0,\"feature_layer\":-2,\"learning_rate\":0.01,\"max_iter\":200,\"htb_batch_size\":32,\"poison_fraction\":0.10,\"nb_epochs\":8,\"seed\":42}", "type": "text" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } } ] } ] }, { "name": "Image", "description": "Image attacks — built-in MNIST / CIFAR-10 or a class-per-folder ZIP.", "item": [ { "name": "Evasion", "description": "Adversarial inputs that fool the model at inference time.", "item": [ { "name": "White-box (gradient access)", "item": [ { "name": "FGSM", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/whitebox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "whitebox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "fgsm", "type": "text" }, { "key": "params", "value": "{\"eps\":0.03}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } }, { "name": "PGD", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/whitebox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "whitebox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "pgd", "type": "text" }, { "key": "params", "value": "{\"eps\":0.03,\"eps_step\":0.01,\"max_iter\":40}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } }, { "name": "BIM", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/whitebox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "whitebox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "bim", "type": "text" }, { "key": "params", "value": "{\"eps\":0.03,\"eps_step\":0.01,\"max_iter\":10}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } }, { "name": "DeepFool", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/whitebox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "whitebox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "deepfool", "type": "text" }, { "key": "params", "value": "{\"max_iter\":50}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } }, { "name": "JSMA", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/whitebox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "whitebox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "jsma", "type": "text" }, { "key": "params", "value": "{\"theta\":0.1,\"gamma\":0.1}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } }, { "name": "APGD", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/whitebox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "whitebox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "apgd", "type": "text" }, { "key": "params", "value": "{\"eps\":0.03,\"max_iter\":40}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } }, { "name": "CW", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/whitebox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "whitebox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "cw", "type": "text" }, { "key": "params", "value": "{\"confidence\":0.0,\"learning_rate\":0.01,\"binary_search_steps\":5,\"max_iter\":20}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } } ] }, { "name": "Black-box (decision-only)", "item": [ { "name": "ZOO", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/blackbox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "blackbox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "zoo", "type": "text" }, { "key": "params", "value": "{\"max_iter\":50,\"nb_parallel\":10}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } }, { "name": "Boundary", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/blackbox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "blackbox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "boundary", "type": "text" }, { "key": "params", "value": "{\"max_iter\":2500,\"epsilon\":0.01}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } }, { "name": "HopSkipJump", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/blackbox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "blackbox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "hsj", "type": "text" }, { "key": "params", "value": "{\"max_iter\":20,\"max_eval\":10000}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } } ] }, { "name": "Grey-box (gradients on surrogate)", "item": [ { "name": "FGSM", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/greybox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "greybox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "fgsm", "type": "text" }, { "key": "params", "value": "{\"eps\":0.03}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } }, { "name": "PGD", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/greybox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "greybox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "pgd", "type": "text" }, { "key": "params", "value": "{\"eps\":0.03,\"eps_step\":0.01,\"max_iter\":40}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } }, { "name": "BIM", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/greybox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "greybox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "bim", "type": "text" }, { "key": "params", "value": "{\"eps\":0.03,\"eps_step\":0.01,\"max_iter\":10}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } }, { "name": "DeepFool", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/greybox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "greybox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "deepfool", "type": "text" }, { "key": "params", "value": "{\"max_iter\":50}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } }, { "name": "JSMA", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/greybox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "greybox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "jsma", "type": "text" }, { "key": "params", "value": "{\"theta\":0.1,\"gamma\":0.1}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } }, { "name": "APGD", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/greybox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "greybox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "apgd", "type": "text" }, { "key": "params", "value": "{\"eps\":0.03,\"max_iter\":40}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } }, { "name": "CW", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/evasion/greybox", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "evasion", "greybox" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "cw", "type": "text" }, { "key": "params", "value": "{\"confidence\":0.0,\"learning_rate\":0.01,\"binary_search_steps\":5,\"max_iter\":20}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } } ] } ] }, { "name": "Inference (Membership)", "description": "Test whether the model leaks its training-set membership.", "item": [ { "name": "MIA · Random Forest meta-classifier", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/inference", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "inference" ] }, "description": "Rule-Based + Black-Box MIA run in parallel. Change `attack_model_type` to gb or lr to swap the meta-classifier.", "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack_model_type", "value": "rf", "type": "text", "description": "rf · gb · lr" }, { "key": "params", "value": "{\"split_ratio\":0.7,\"max_samples\":500,\"seed\":42}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } } ] }, { "name": "Extraction (CopycatCNN)", "description": "Train a stolen ('thieved') copy of the model from victim queries.", "item": [ { "name": "CopycatCNN", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/extraction", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "extraction" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "params", "value": "{\"nb_stolen\":200,\"nb_epochs\":3,\"batch_size\":32,\"use_probability\":true}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } } ] }, { "name": "Poisoning", "description": "Inject malicious training samples, retrain a clone, measure ASR.", "item": [ { "name": "Backdoor", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/poisoning", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "poisoning" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "backdoor", "type": "text" }, { "key": "params", "value": "{\"target_class\":0,\"patch_size\":4,\"patch_value\":1.0,\"patch_position\":\"bottom-right\",\"poison_fraction\":0.10,\"nb_epochs\":5,\"batch_size\":32,\"seed\":42}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } }, { "name": "Clean-Label", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/poisoning", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "poisoning" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "clean_label", "type": "text" }, { "key": "params", "value": "{\"target_class\":0,\"patch_size\":4,\"patch_value\":1.0,\"patch_position\":\"bottom-right\",\"poison_fraction\":0.10,\"pp_poison\":0.33,\"eps\":0.3,\"eps_step\":0.1,\"max_iter\":100,\"nb_epochs\":5,\"batch_size\":32,\"seed\":42}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } }, { "name": "Hidden Trigger", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/poisoning", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "poisoning" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "image", "type": "text" }, { "key": "attack", "value": "hidden_trigger", "type": "text" }, { "key": "params", "value": "{\"target_class\":0,\"source_class\":1,\"patch_size\":4,\"patch_value\":1.0,\"patch_position\":\"bottom-right\",\"feature_layer\":-2,\"learning_rate\":0.01,\"max_iter\":200,\"htb_batch_size\":32,\"poison_fraction\":0.10,\"nb_epochs\":5,\"seed\":42}", "type": "text" }, { "key": "source", "value": "mnist", "type": "text", "description": "mnist · cifar10 · zip" }, { "key": "limit", "value": "200", "type": "text", "description": "Max samples to use" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5) — optional for extraction" } ] } } } ] } ] }, { "name": "Adversarial Penetration Test", "description": "ε-sweep across FGSM / PGD / BIM with a pass/fail verdict. Structured data only.", "item": [ { "name": "Full sweep (ZIP download — default)", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/pentest", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "pentest" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text", "description": "Adversarial Penetration Test is structured-only." }, { "key": "use_fgsm", "value": "true", "type": "text" }, { "key": "use_pgd", "value": "true", "type": "text" }, { "key": "use_bim", "value": "true", "type": "text" }, { "key": "params", "value": "{\"eps_step\":0.05}", "type": "text", "description": "Optional STIX metadata: model_id, dataset, norm" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } }, { "name": "Full sweep (JSON envelope)", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/pentest?format=json", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "pentest" ], "query": [ { "key": "format", "value": "json" } ] }, "body": { "mode": "formdata", "formdata": [ { "key": "data_type", "value": "structured", "type": "text", "description": "Adversarial Penetration Test is structured-only." }, { "key": "use_fgsm", "value": "true", "type": "text" }, { "key": "use_pgd", "value": "true", "type": "text" }, { "key": "use_bim", "value": "true", "type": "text" }, { "key": "params", "value": "{\"eps_step\":0.05}", "type": "text", "description": "Optional STIX metadata: model_id, dataset, norm" }, { "key": "columns_to_drop", "value": "flow_id,src_ip,dst_ip,src_port,dst_port", "type": "text", "description": "Comma-separated columns to drop before scaling" }, { "key": "label_col", "value": "label", "type": "text", "description": "Ground-truth column" }, { "key": "label_binary_col", "value": "", "type": "text", "description": "Optional binary label column" }, { "key": "dataset", "type": "file", "src": [], "description": "Tabular dataset (.csv)" }, { "key": "model", "type": "file", "src": [], "description": "Keras model (.h5)" }, { "key": "scaler", "type": "file", "src": [], "description": "Fitted scaler (.joblib / .pkl)" }, { "key": "encoder", "type": "file", "src": [], "description": "Label encoder (.joblib / .pkl)" } ] } } } ] }, { "name": "Object detection", "description": "ART evasion attacks against object detectors (Faster R-CNN, YOLOv5) on a built-in COCO sample or your own JPEG/PNG images. Start with *Catalog* (models, attacks + parameter ranges, compatibility, sample image ids). Same job contract as every family: slow runs return `202` — use the Jobs folder.", "item": [ { "name": "Catalog (models · attacks · compatibility · sample images)", "request": { "method": "GET", "header": [], "url": { "raw": "{{baseUrl}}/api/object-detection/catalog", "host": [ "{{baseUrl}}" ], "path": [ "api", "object-detection", "catalog" ] } } }, { "name": "Sample image (thumbnail)", "request": { "method": "GET", "header": [], "url": { "raw": "{{baseUrl}}/api/object-detection/images/000000206487.jpg?thumb=true", "host": [ "{{baseUrl}}" ], "path": [ "api", "object-detection", "images", "000000206487.jpg" ], "query": [ { "key": "thumb", "value": "true" } ] } } }, { "name": "FGSM (L∞) (ZIP download — default)", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/object-detection", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "object-detection" ] }, "description": "One-step L∞ sign-gradient perturbation of the whole image that increases the detector's training loss. Cheap baseline; an untargeted loss increase does not guarantee that any particular object disappears.", "body": { "mode": "formdata", "formdata": [ { "key": "model", "value": "yolov5s", "type": "text", "description": "Built-in detector: frcnn_mobilenet_v3_320 · frcnn_resnet50_fpn · yolov5s" }, { "key": "attack", "value": "fgsm", "type": "text" }, { "key": "params", "value": "{\"eps\": 8}", "type": "text", "description": "eps / eps_step / learning_rate are in 0–255 pixel units on every model (8 = 8/255 on [0,1] models)." }, { "key": "images", "value": "000000206487.jpg,000000118209.jpg", "type": "text", "description": "Built-in COCO sample ids (see the catalog). Empty + no uploads → first 4 samples." }, { "key": "label_source", "value": "ground_truth", "type": "text", "description": "ground_truth · clean_predictions" }, { "key": "score_threshold", "value": "0.5", "type": "text" }, { "key": "seed", "value": "0", "type": "text" } ] } } }, { "name": "PGD (L∞) (ZIP download — default)", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/object-detection", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "object-detection" ] }, "description": "Iterative L∞ projected gradient descent on the detection loss — the attack in ART's official YOLOv3/v5 example. Stronger than FGSM at the same budget.", "body": { "mode": "formdata", "formdata": [ { "key": "model", "value": "yolov5s", "type": "text", "description": "Built-in detector: frcnn_mobilenet_v3_320 · frcnn_resnet50_fpn · yolov5s" }, { "key": "attack", "value": "pgd", "type": "text" }, { "key": "params", "value": "{\"eps\": 8, \"eps_step\": 2, \"max_iter\": 10}", "type": "text", "description": "eps / eps_step / learning_rate are in 0–255 pixel units on every model (8 = 8/255 on [0,1] models)." }, { "key": "images", "value": "000000206487.jpg,000000118209.jpg", "type": "text", "description": "Built-in COCO sample ids (see the catalog). Empty + no uploads → first 4 samples." }, { "key": "label_source", "value": "ground_truth", "type": "text", "description": "ground_truth · clean_predictions" }, { "key": "score_threshold", "value": "0.5", "type": "text" }, { "key": "seed", "value": "0", "type": "text" } ] } } }, { "name": "DPatch (ZIP download — default)", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/object-detection", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "object-detection" ] }, "description": "Learns one universal rectangular patch (shared across all images) that disrupts the detector; pasted at the top-left corner. Unbounded inside the patch, zero change outside.", "body": { "mode": "formdata", "formdata": [ { "key": "model", "value": "yolov5s", "type": "text", "description": "Built-in detector: frcnn_mobilenet_v3_320 · frcnn_resnet50_fpn · yolov5s" }, { "key": "attack", "value": "dpatch", "type": "text" }, { "key": "params", "value": "{\"patch_size\": 64, \"learning_rate\": 8, \"max_iter\": 40}", "type": "text", "description": "eps / eps_step / learning_rate are in 0–255 pixel units on every model (8 = 8/255 on [0,1] models)." }, { "key": "images", "value": "000000206487.jpg,000000118209.jpg", "type": "text", "description": "Built-in COCO sample ids (see the catalog). Empty + no uploads → first 4 samples." }, { "key": "label_source", "value": "ground_truth", "type": "text", "description": "ground_truth · clean_predictions" }, { "key": "score_threshold", "value": "0.5", "type": "text" }, { "key": "seed", "value": "0", "type": "text" } ] } } }, { "name": "RobustDPatch (ZIP download — default)", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/object-detection", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "object-detection" ] }, "description": "DPatch with expectation-over-transformation: the patch is optimised under random crops, brightness changes and (optionally) 90° rotations so it survives mild image changes. Not a claim of physical-world robustness.", "body": { "mode": "formdata", "formdata": [ { "key": "model", "value": "yolov5s", "type": "text", "description": "Built-in detector: frcnn_mobilenet_v3_320 · frcnn_resnet50_fpn · yolov5s" }, { "key": "attack", "value": "robust_dpatch", "type": "text" }, { "key": "params", "value": "{\"patch_size\": 64, \"patch_x\": 16, \"patch_y\": 16, \"learning_rate\": 8, \"max_iter\": 30, \"sample_size\": 2, \"brightness\": 0.2, \"crop\": 8, \"rotations\": false}", "type": "text", "description": "eps / eps_step / learning_rate are in 0–255 pixel units on every model (8 = 8/255 on [0,1] models)." }, { "key": "images", "value": "000000206487.jpg,000000118209.jpg", "type": "text", "description": "Built-in COCO sample ids (see the catalog). Empty + no uploads → first 4 samples." }, { "key": "label_source", "value": "ground_truth", "type": "text", "description": "ground_truth · clean_predictions" }, { "key": "score_threshold", "value": "0.5", "type": "text" }, { "key": "seed", "value": "0", "type": "text" } ] } } }, { "name": "Adversarial Patch (PyTorch) (ZIP download — default)", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/object-detection", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "object-detection" ] }, "description": "Learns a square patch with Adam under random rotation/scale; applied at a fixed location. Untargeted: maximises the detector's loss against its own clean predictions.", "body": { "mode": "formdata", "formdata": [ { "key": "model", "value": "yolov5s", "type": "text", "description": "Built-in detector: frcnn_mobilenet_v3_320 · frcnn_resnet50_fpn · yolov5s" }, { "key": "attack", "value": "adv_patch", "type": "text" }, { "key": "params", "value": "{\"patch_size\": 64, \"patch_x\": 16, \"patch_y\": 16, \"learning_rate\": 8, \"max_iter\": 40, \"rotation_max\": 10}", "type": "text", "description": "eps / eps_step / learning_rate are in 0–255 pixel units on every model (8 = 8/255 on [0,1] models)." }, { "key": "images", "value": "000000206487.jpg,000000118209.jpg", "type": "text", "description": "Built-in COCO sample ids (see the catalog). Empty + no uploads → first 4 samples." }, { "key": "label_source", "value": "ground_truth", "type": "text", "description": "ground_truth · clean_predictions" }, { "key": "score_threshold", "value": "0.5", "type": "text" }, { "key": "seed", "value": "0", "type": "text" } ] } } }, { "name": "PGD (JSON envelope)", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/object-detection?format=json", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "object-detection" ], "query": [ { "key": "format", "value": "json" } ] }, "body": { "mode": "formdata", "formdata": [ { "key": "model", "value": "yolov5s", "type": "text", "description": "Built-in detector: frcnn_mobilenet_v3_320 · frcnn_resnet50_fpn · yolov5s" }, { "key": "attack", "value": "pgd", "type": "text" }, { "key": "params", "value": "{\"eps\": 8, \"eps_step\": 2, \"max_iter\": 10}", "type": "text", "description": "eps / eps_step / learning_rate are in 0–255 pixel units on every model (8 = 8/255 on [0,1] models)." }, { "key": "images", "value": "000000206487.jpg,000000118209.jpg", "type": "text", "description": "Built-in COCO sample ids (see the catalog). Empty + no uploads → first 4 samples." }, { "key": "label_source", "value": "ground_truth", "type": "text", "description": "ground_truth · clean_predictions" }, { "key": "score_threshold", "value": "0.5", "type": "text" }, { "key": "seed", "value": "0", "type": "text" } ] } } }, { "name": "PGD on your own images (attach files)", "request": { "method": "POST", "header": [], "url": { "raw": "{{baseUrl}}/api/attack/object-detection", "host": [ "{{baseUrl}}" ], "path": [ "api", "attack", "object-detection" ] }, "body": { "mode": "formdata", "formdata": [ { "key": "model", "value": "yolov5s", "type": "text", "description": "Built-in detector: frcnn_mobilenet_v3_320 · frcnn_resnet50_fpn · yolov5s" }, { "key": "attack", "value": "pgd", "type": "text" }, { "key": "params", "value": "{\"eps\": 8, \"eps_step\": 2, \"max_iter\": 10}", "type": "text", "description": "eps / eps_step / learning_rate are in 0–255 pixel units on every model (8 = 8/255 on [0,1] models)." }, { "key": "label_source", "value": "ground_truth", "type": "text", "description": "ground_truth · clean_predictions" }, { "key": "score_threshold", "value": "0.5", "type": "text" }, { "key": "seed", "value": "0", "type": "text" }, { "key": "images", "value": "", "type": "text", "description": "Leave empty to attack only the uploaded images." }, { "key": "files", "type": "file", "src": [], "description": "Your JPEG/PNG image(s) — repeat the `files` key for several." } ] } } } ] }, { "name": "Jobs (async attacks)", "description": "If an attack endpoint returns `202` instead of your result, copy its `job_id` into the `jobId` collection variable and use these requests to poll status, fetch the finished result, or cancel it.", "item": [ { "name": "Poll job status", "request": { "method": "GET", "header": [], "url": { "raw": "{{baseUrl}}/api/jobs/{{jobId}}", "host": [ "{{baseUrl}}" ], "path": [ "api", "jobs", "{{jobId}}" ] } } }, { "name": "Fetch job result (ZIP)", "request": { "method": "GET", "header": [], "url": { "raw": "{{baseUrl}}/api/jobs/{{jobId}}/result?format=zip", "host": [ "{{baseUrl}}" ], "path": [ "api", "jobs", "{{jobId}}", "result" ], "query": [ { "key": "format", "value": "zip" } ] } } }, { "name": "Fetch job result (JSON)", "request": { "method": "GET", "header": [], "url": { "raw": "{{baseUrl}}/api/jobs/{{jobId}}/result?format=json", "host": [ "{{baseUrl}}" ], "path": [ "api", "jobs", "{{jobId}}", "result" ], "query": [ { "key": "format", "value": "json" } ] } } }, { "name": "Cancel job", "request": { "method": "DELETE", "header": [], "url": { "raw": "{{baseUrl}}/api/jobs/{{jobId}}", "host": [ "{{baseUrl}}" ], "path": [ "api", "jobs", "{{jobId}}" ] } } } ] } ], "variable": [ { "key": "baseUrl", "value": "https://127.0.0.1:8000", "type": "string", "description": "API root. Override per-environment if you're hitting a remote deployment. This deployment uses a self-signed cert (no domain points at the server yet) — in Postman, turn off Settings → General → 'SSL certificate verification' or you'll get a self-signed-cert error on every request." }, { "key": "apiKey", "value": "", "type": "string", "description": "Your X-API-Key. Every request in this collection sends it automatically once set." }, { "key": "jobId", "value": "", "type": "string", "description": "Paste a job_id here after an attack request returns 202 (see the Jobs folder)." } ], "auth": { "type": "apikey", "apikey": [ { "key": "key", "value": "X-API-Key", "type": "string" }, { "key": "value", "value": "{{apiKey}}", "type": "string" }, { "key": "in", "value": "header", "type": "string" } ] } }