File size: 11,898 Bytes
50f71a7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
87f2d84
 
 
 
 
 
 
50f71a7
87f2d84
 
 
 
 
50f71a7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
"""Synthetic episode generator for the Fake Gang Detection environment."""

from __future__ import annotations

import json
import math
import random
import uuid
from pathlib import Path
from typing import Any, Dict, List, Tuple

# ---------------------------------------------------------------------------
# Task configuration
# ---------------------------------------------------------------------------

TASK_CONFIG = {
    "easy": {
        "network_size": 50,
        "gang_size": 10,
        "decoy_count": 0,
        "max_steps": 30,
        "starting_visible": 5,
        "post_hour_mean": 14.0,
        "post_hour_std": 0.5,
        "photo_reuse_range": (0.70, 0.95),
        "bio_template_range": (0.60, 0.90),
        "intra_gang_density": 0.80,
        "evasion_schedule": [],
        "win_recall": 0.8,
        "win_precision": 0.7,
    },
    "medium": {
        "network_size": 200,
        "gang_size": 10,
        "decoy_count": 20,
        "max_steps": 50,
        "starting_visible": 10,
        "post_hour_mean": 14.0,
        "post_hour_std": 1.5,
        "photo_reuse_range": (0.40, 0.70),
        "bio_template_range": (0.30, 0.60),
        "intra_gang_density": 0.70,
        "evasion_schedule": [
            {"step": 20, "event": "unfollow_intragang", "drop_rate": 0.5}
        ],
        "win_recall": 0.8,
        "win_precision": 0.7,
    },
    "hard": {
        "network_size": 1000,
        "gang_size": 10,
        "decoy_count": 50,
        "max_steps": 80,
        "starting_visible": 20,
        "post_hour_mean": 14.0,
        "post_hour_std": 2.5,
        "photo_reuse_range": (0.30, 0.55),
        "bio_template_range": (0.20, 0.50),
        "intra_gang_density": 0.60,
        "evasion_schedule": [
            {"step": 15, "event": "unfollow_intragang", "drop_rate": 0.3, "rename_count": 3},
            {"step": 30, "event": "unfollow_intragang", "drop_rate": 0.3, "rename_count": 3},
            {"step": 45, "event": "unfollow_intragang", "drop_rate": 0.3, "rename_count": 3},
            {"step": 60, "event": "unfollow_intragang", "drop_rate": 0.3, "rename_count": 3},
        ],
        "win_recall": 0.9,
        "win_precision": 0.8,
    },
}


# ---------------------------------------------------------------------------
# Helpers
# ---------------------------------------------------------------------------

def _lognormal_int(rng: random.Random, mu: float, sigma: float, lo: int, hi: int) -> int:
    val = math.exp(rng.gauss(mu, sigma))
    return max(lo, min(hi, int(val)))


def _beta(rng: random.Random, a: float, b: float) -> float:
    """Simple beta distribution via gamma approximation."""
    x = rng.gammavariate(a, 1)
    y = rng.gammavariate(b, 1)
    return x / (x + y)


# ---------------------------------------------------------------------------
# Account generators
# ---------------------------------------------------------------------------

def _gen_real_account(rng: random.Random, acc_id: str) -> Dict[str, Any]:
    return {
        "id": acc_id,
        "is_fake": False,
        "gang_id": None,
        "features": {
            "follower_count": _lognormal_int(rng, 6.0, 1.5, 10, 100_000),
            "following_count": _lognormal_int(rng, 5.5, 1.2, 10, 5_000),
            "post_count": _lognormal_int(rng, 4.0, 1.0, 1, 5_000),
            "avg_post_hour": rng.uniform(6, 23),
            "photo_reuse_score": round(_beta(rng, 1, 10), 4),
            "bio_template_score": round(_beta(rng, 1, 8), 4),
            "account_age_days": rng.randint(30, 1825),
            "comment_repeat_score": round(_beta(rng, 1, 20), 4),  # mostly 0.0-0.08
            "ip_cluster_id": f"ip_real_{acc_id}",
            "shared_ip_count": 0,
            "name_change_count": 0,
        },
        "true_edges": {"follows": [], "followed_by": []},
    }


def _gen_decoy_account(rng: random.Random, acc_id: str) -> Dict[str, Any]:
    acc = _gen_real_account(rng, acc_id)
    acc["features"]["photo_reuse_score"] = round(rng.uniform(0.2, 0.4), 4)
    acc["features"]["bio_template_score"] = round(rng.uniform(0.2, 0.4), 4)
    acc["features"]["comment_repeat_score"] = round(rng.uniform(0.10, 0.30), 4)
    return acc


def _gen_celebrity_account(rng: random.Random, acc_id: str) -> Dict[str, Any]:
    return {
        "id": acc_id,
        "is_fake": False,
        "gang_id": None,
        "features": {
            "follower_count": rng.randint(100_000, 5_000_000),
            "following_count": rng.randint(50, 500),
            "post_count": rng.randint(500, 5000),
            "avg_post_hour": rng.uniform(6, 23),
            "photo_reuse_score": round(rng.uniform(0.0, 0.05), 4),
            "bio_template_score": round(rng.uniform(0.0, 0.05), 4),
            "account_age_days": rng.randint(1000, 3000),
            "comment_repeat_score": round(rng.uniform(0.0, 0.03), 4),
            "ip_cluster_id": f"ip_celeb_{acc_id}",
            "shared_ip_count": 0,
            "name_change_count": 0,
        },
        "true_edges": {"follows": [], "followed_by": []},
    }


def _gen_gang_accounts(
    rng: random.Random,
    ids: List[str],
    cfg: Dict[str, Any],
    seed: int,
) -> List[Dict[str, Any]]:
    gang_size = len(ids)
    base_age = rng.randint(30, 180)
    base_hour = cfg["post_hour_mean"]
    ph_lo, ph_hi = cfg["photo_reuse_range"]
    bt_lo, bt_hi = cfg["bio_template_range"]
    std = cfg["post_hour_std"]
    accounts = []
    for acc_id in ids:
        accounts.append({
            "id": acc_id,
            "is_fake": True,
            "gang_id": "gang_A",
            "features": {
                "follower_count": rng.randint(1_000, 8_000),
                "following_count": rng.randint(200, 2_000),
                "post_count": rng.randint(50, 500),
                "avg_post_hour": round(max(0, min(23, rng.gauss(base_hour, std))), 2),
                "photo_reuse_score": round(rng.uniform(ph_lo, ph_hi), 4),
                "bio_template_score": round(rng.uniform(bt_lo, bt_hi), 4),
                "account_age_days": base_age + rng.randint(0, 7),
                "comment_repeat_score": round(rng.uniform(0.60, 0.90), 4),
                "ip_cluster_id": f"ip_gang_{seed}",
                "shared_ip_count": gang_size - 1,
                "name_change_count": 0,
            },
            "true_edges": {"follows": [], "followed_by": []},
        })
    return accounts


# ---------------------------------------------------------------------------
# Edge generation
# ---------------------------------------------------------------------------

def _build_edges(
    rng: random.Random,
    accounts: List[Dict[str, Any]],
    gang_ids: List[str],
    density: float,
) -> None:
    """Add follower edges in-place."""
    acc_map = {a["id"]: a for a in accounts}
    all_ids = [a["id"] for a in accounts]

    # Intra-gang edges (high density)
    for g in gang_ids:
        for h in gang_ids:
            if g != h and rng.random() < density:
                acc_map[g]["true_edges"]["follows"].append(h)
                acc_map[h]["true_edges"]["followed_by"].append(g)

    # Real/decoy accounts: sparse preferential attachment
    non_gang = [a["id"] for a in accounts if a["id"] not in gang_ids]
    for acc_id in non_gang:
        n_follows = rng.randint(5, min(50, len(all_ids) - 1))
        targets = rng.sample([x for x in all_ids if x != acc_id], min(n_follows, len(all_ids) - 1))
        for t in targets:
            acc_map[acc_id]["true_edges"]["follows"].append(t)
            acc_map[t]["true_edges"]["followed_by"].append(acc_id)


# ---------------------------------------------------------------------------
# Episode generator
# ---------------------------------------------------------------------------

def generate_episode(task: str, seed: int) -> Dict[str, Any]:
    cfg = TASK_CONFIG[task]
    rng = random.Random(seed)

    network_size = cfg["network_size"]
    gang_size = cfg["gang_size"]
    decoy_count = cfg["decoy_count"]
    real_count = network_size - gang_size - decoy_count

    # Generate IDs
    all_ids = [f"acc_{i:04d}" for i in range(network_size)]
    rng.shuffle(all_ids)
    gang_ids = all_ids[:gang_size]
    decoy_ids = all_ids[gang_size: gang_size + decoy_count]
    real_ids = all_ids[gang_size + decoy_count:]

    # Partition real IDs: last 2 → celebrities, 2 before that → zero-edge isolates
    # Need at least 4 real accounts for this split; fall back gracefully if not.
    if len(real_ids) >= 4:
        celeb_ids = real_ids[-2:]
        zero_edge_ids = real_ids[-4:-2]
        standard_real_ids = real_ids[:-4]
    elif len(real_ids) >= 2:
        celeb_ids = real_ids[-2:]
        zero_edge_ids = []
        standard_real_ids = real_ids[:-2]
    else:
        celeb_ids = []
        zero_edge_ids = []
        standard_real_ids = real_ids

    accounts: List[Dict[str, Any]] = []
    accounts.extend(_gen_gang_accounts(rng, gang_ids, cfg, seed))
    accounts.extend(_gen_decoy_account(rng, did) for did in decoy_ids)
    accounts.extend(_gen_real_account(rng, rid) for rid in standard_real_ids)

    # Zero-edge isolates: real accounts with no connections (edge cases for the agent)
    for rid in zero_edge_ids:
        acc = _gen_real_account(rng, rid)
        acc["features"]["follower_count"] = 0
        acc["features"]["following_count"] = 0
        accounts.append(acc)

    # Celebrity accounts: large legit hubs (high hub_legitimacy → low fake_risk)
    for cid in celeb_ids:
        accounts.append(_gen_celebrity_account(rng, cid))

    _build_edges(rng, accounts, gang_ids, cfg["intra_gang_density"])

    # Choose starting visible accounts.
    # Guarantee exactly 1 gang member is included so the cascade CAN start
    # regardless of seed. The agent still has to identify WHICH account is fake
    # (requires inspecting profiles) — so difficulty is preserved.
    # Without this, ~31% of easy episodes and ~82% of hard episodes start with
    # zero gang members visible, making score variance seed-luck rather than
    # agent skill.
    starting_count = cfg["starting_visible"]
    forced_gang = rng.sample(gang_ids, 1)          # exactly 1 gang member
    rest_pool = [i for i in all_ids if i not in forced_gang]
    additional = rng.sample(rest_pool, starting_count - 1)
    starting_visible = forced_gang + additional
    rng.shuffle(starting_visible)                   # don't reveal which is fake

    return {
        "episode_id": str(uuid.uuid4()),
        "task": task,
        "seed": seed,
        "max_steps": cfg["max_steps"],
        "win_recall": cfg["win_recall"],
        "win_precision": cfg["win_precision"],
        "starting_visible": starting_visible,
        "gang_member_ids": gang_ids,
        "decoy_ids": decoy_ids,
        "celeb_ids": celeb_ids,
        "zero_edge_ids": zero_edge_ids,
        "network": {"accounts": accounts},
        "evasion_schedule": cfg["evasion_schedule"],
    }


# ---------------------------------------------------------------------------
# Bulk generation
# ---------------------------------------------------------------------------

def generate_all_episodes(
    output_dir: Path,
    counts: Dict[str, int] | None = None,
) -> None:
    if counts is None:
        counts = {"easy": 50, "medium": 50, "hard": 50}

    output_dir.mkdir(parents=True, exist_ok=True)
    total = 0
    for task, n in counts.items():
        for i in range(n):
            seed = i  # deterministic
            ep = generate_episode(task, seed)
            fname = output_dir / f"{task}_{i:03d}.json"
            fname.write_text(json.dumps(ep, indent=2))
            total += 1
    print(f"Generated {total} episodes in {output_dir}")


def main() -> None:
    """Entry point: called by `uv run generate-episodes` or directly."""
    here = Path(__file__).parent.parent / "episodes"
    generate_all_episodes(here)


if __name__ == "__main__":
    main()