xiaohy commited on
Commit
687782e
·
verified ·
1 Parent(s): db0cc64

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +62 -129
app.py CHANGED
@@ -1,6 +1,6 @@
1
  # ================================================================
2
- # 教育大模型MIA攻防研究 - Gradio演示系统 v6.1 Final (苹果风)
3
- # 整合了双雷达图 + 算法流程图 + 伪代码 + 详尽据分析 + 完整结论
4
  # ================================================================
5
 
6
  import os
@@ -31,7 +31,6 @@ def clean_text(text):
31
  text = re.sub(r'[\u200b-\u200f\u2028-\u202f\u2060-\u206f\ufeff]', '', text)
32
  return text.strip()
33
 
34
- # 尝试加载数据,如果不存在则使用虚拟数据以确保运行
35
  try:
36
  member_data = load_json("data/member.json")
37
  non_member_data = load_json("data/non_member.json")
@@ -58,12 +57,11 @@ except FileNotFoundError:
58
  for s in [0.005, 0.01, 0.015, 0.02, 0.025, 0.03]:
59
  k = f"perturbation_{s}"
60
  perturb_results[k] = {m: v*0.85 for m, v in mia_results["baseline"].items()}
61
- # 模拟方差变大
62
  perturb_results[k]["member_loss_std"] = np.sqrt(0.03**2 + s**2)
63
  perturb_results[k]["non_member_loss_std"] = np.sqrt(0.03**2 + s**2)
64
 
65
  # ================================================================
66
- # 全局图表配置 - 简约苹果风
67
  # ================================================================
68
  COLORS = {
69
  'bg': '#FFFFFF',
@@ -80,8 +78,6 @@ COLORS = {
80
  'ls_colors': ['#A0C4FF', '#70A1FF', '#478EFF', '#007AFF'],
81
  'op_colors': ['#98F5E1', '#6EE7B7', '#34D399', '#10B981', '#059669', '#047857'],
82
  }
83
-
84
- # 图表宽度配置 (为了适配双雷达图)
85
  CHART_W = 14
86
 
87
  def apply_light_style(fig, ax_or_axes):
@@ -100,17 +96,15 @@ def apply_light_style(fig, ax_or_axes):
100
  ax.grid(True, color=COLORS['grid'], alpha=0.6, linestyle='-', linewidth=0.8)
101
  ax.set_axisbelow(True)
102
 
103
- # ================================================================
104
- # 提取指标的辅助函数
105
- # ================================================================
106
  LS_KEYS = ["baseline", "smooth_eps_0.02", "smooth_eps_0.05", "smooth_eps_0.1", "smooth_eps_0.2"]
107
- LS_LABELS_EN = ["Baseline", "LS(e=0.02)", "LS(e=0.05)", "LS(e=0.1)", "LS(e=0.2)"]
108
- LS_LABELS_CN = ["基线", "LS(ε=0.02)", "LS(ε=0.05)", "LS(ε=0.1)", "LS(ε=0.2)"]
109
 
110
  OP_SIGMAS = [0.005, 0.01, 0.015, 0.02, 0.025, 0.03]
111
  OP_KEYS = [f"perturbation_{s}" for s in OP_SIGMAS]
112
- OP_LABELS_EN = [f"OP(s={s})" for s in OP_SIGMAS]
113
- OP_LABELS_CN = [f"OP(σ={s})" for s in OP_SIGMAS]
114
 
115
  ALL_KEYS = LS_KEYS + OP_KEYS
116
 
@@ -129,12 +123,8 @@ bl_acc = gu("baseline")
129
  bl_m_mean = gm("baseline", "member_loss_mean")
130
  bl_nm_mean = gm("baseline", "non_member_loss_mean")
131
 
132
- TYPE_CN = {'calculation': '基础计算', 'word_problem': '应用题',
133
- 'concept': '概念问答', 'error_correction': '错题订正'}
134
 
135
- # ================================================================
136
- # 效用评估题库
137
- # ================================================================
138
  np.random.seed(777)
139
  EVAL_POOL = []
140
  _types = ['calculation']*120 + ['word_problem']*90 + ['concept']*60 + ['error_correction']*30
@@ -148,8 +138,7 @@ for _i in range(300):
148
  else: _q,_ans=f"{_a} x {_b} = ?",str(_a*_b)
149
  elif _t == 'word_problem':
150
  _a,_b = int(np.random.randint(5,200)), int(np.random.randint(3,50))
151
- _tpls = [(f"{_a} apples, ate {_b}, left?",str(_a-_b)),
152
- (f"{_a} per group, {_b} groups, total?",str(_a*_b))]
153
  _q,_ans = _tpls[_i%len(_tpls)]
154
  elif _t == 'concept':
155
  _cs = [("area","Area = space occupied by a shape"),("perimeter","Perimeter = total boundary length")]
@@ -164,49 +153,35 @@ for _i in range(300):
164
  EVAL_POOL.append(item)
165
 
166
  # ================================================================
167
- # 图表绘制函数
168
  # ================================================================
169
  def fig_gauge(loss_val, m_mean, nm_mean, thr, m_std, nm_std):
170
- fig, ax = plt.subplots(figsize=(10, 2.6))
171
- fig.patch.set_facecolor(COLORS['bg'])
172
- ax.set_facecolor(COLORS['panel'])
173
-
174
- xlo = min(m_mean - 3.0 * m_std, loss_val - 0.005)
175
- xhi = max(nm_mean + 3.0 * nm_std, loss_val + 0.005)
176
-
177
- ax.axvspan(xlo, thr, alpha=0.2, color=COLORS['accent'])
178
- ax.axvspan(thr, xhi, alpha=0.2, color=COLORS['danger'])
179
-
180
  ax.axvline(m_mean, color=COLORS['accent'], lw=2, ls=':', alpha=0.8, zorder=2)
181
  ax.text(m_mean - 0.002, 1.02, f'Member Mean\n{m_mean:.4f}', ha='right', va='bottom', fontsize=9, color=COLORS['accent'], transform=ax.get_xaxis_transform())
182
-
183
  ax.axvline(nm_mean, color=COLORS['danger'], lw=2, ls=':', alpha=0.8, zorder=2)
184
  ax.text(nm_mean + 0.002, 1.02, f'Non-Member Mean\n{nm_mean:.4f}', ha='left', va='bottom', fontsize=9, color=COLORS['danger'], transform=ax.get_xaxis_transform())
185
-
186
  ax.axvline(thr, color=COLORS['text_dim'], lw=2.5, ls='--', zorder=3)
187
  ax.text(thr, 1.25, f'Threshold\n{thr:.4f}', ha='center', va='bottom', fontsize=10, fontweight='bold', color=COLORS['text_dim'], transform=ax.get_xaxis_transform())
188
-
189
  mc = COLORS['accent'] if loss_val < thr else COLORS['danger']
190
  ax.plot(loss_val, 0.5, marker='o', ms=16, color='white', mec=mc, mew=3, zorder=5, transform=ax.get_xaxis_transform())
191
  ax.text(loss_val, 0.75, f'Current Loss\n{loss_val:.4f}', ha='center', fontsize=11, fontweight='bold', color=mc, transform=ax.get_xaxis_transform())
192
-
193
  ax.text((xlo+thr)/2, 0.25, 'MEMBER', ha='center', fontsize=12, color=COLORS['accent'], alpha=0.6, fontweight='bold', transform=ax.get_xaxis_transform())
194
  ax.text((thr+xhi)/2, 0.25, 'NON-MEMBER', ha='center', fontsize=12, color=COLORS['danger'], alpha=0.6, fontweight='bold', transform=ax.get_xaxis_transform())
195
-
196
  ax.set_xlim(xlo, xhi); ax.set_yticks([])
197
  for s in ax.spines.values(): s.set_visible(False)
198
- ax.spines['bottom'].set_visible(True); ax.spines['bottom'].set_color(COLORS['grid'])
199
- ax.tick_params(colors=COLORS['text_dim'], width=1)
200
- ax.set_xlabel('Loss Value', fontsize=11, color=COLORS['text'], fontweight='medium')
201
- plt.tight_layout(pad=0.5)
202
  return fig
203
 
204
  def fig_auc_bar():
205
  names, vals, clrs = [], [], []
206
  ls_c = [COLORS['baseline']] + COLORS['ls_colors']
207
- for i,(k,l) in enumerate(zip(LS_KEYS, LS_LABELS_EN)):
208
  if k in mia_results: names.append(l); vals.append(mia_results[k]['auc']); clrs.append(ls_c[i])
209
- for i,(k,l) in enumerate(zip(OP_KEYS, OP_LABELS_EN)):
210
  if k in perturb_results: names.append(l); vals.append(perturb_results[k]['auc']); clrs.append(COLORS['op_colors'][i])
211
  fig, ax = plt.subplots(figsize=(14, 6)); apply_light_style(fig, ax)
212
  bars = ax.bar(range(len(names)), vals, color=clrs, width=0.65, edgecolor='none', zorder=3)
@@ -214,78 +189,35 @@ def fig_auc_bar():
214
  ax.axhline(0.5, color=COLORS['text_dim'], ls='--', lw=1.5, alpha=0.6, label='Random Guess (0.5)', zorder=2)
215
  ax.axhline(bl_auc, color=COLORS['danger'], ls=':', lw=1.5, alpha=0.8, label=f'Baseline ({bl_auc:.4f})', zorder=2)
216
  ax.set_ylabel('MIA Attack AUC', fontsize=12, fontweight='medium'); ax.set_title('Defense Effectiveness: MIA AUC Comparison', fontsize=14, fontweight='bold', pad=20)
217
- ax.set_ylim(0.45, max(vals)+0.05); ax.set_xticks(range(len(names))); ax.set_xticklabels(names, rotation=30, ha='right', fontsize=10)
218
  ax.legend(facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text'], fontsize=10, loc='upper right'); plt.tight_layout()
219
  return fig
220
 
221
  def fig_radar():
222
  ms = ['AUC', 'Atk Acc', 'Prec', 'Recall', 'F1', 'TPR@5%', 'TPR@1%', 'Gap']
223
- mk = ['auc', 'attack_accuracy', 'precision', 'recall', 'f1',
224
- 'tpr_at_5fpr', 'tpr_at_1fpr', 'loss_gap']
225
- N = len(ms)
226
- ag = np.linspace(0, 2 * np.pi, N, endpoint=False).tolist() + [0]
227
-
228
- fig, axes = plt.subplots(1, 2, figsize=(CHART_W + 2, 7),
229
- subplot_kw=dict(polar=True))
230
- fig.patch.set_facecolor('white')
231
-
232
- # --- 左图: 5个标签平滑模型 ---
233
- ls_cfgs = [
234
- ("Baseline", "baseline", '#F04438'),
235
- ("LS(e=0.02)", "smooth_eps_0.02", '#B2DDFF'),
236
- ("LS(e=0.05)", "smooth_eps_0.05", '#84CAFF'),
237
- ("LS(e=0.1)", "smooth_eps_0.1", '#2E90FA'),
238
- ("LS(e=0.2)", "smooth_eps_0.2", '#7A5AF8'),
239
- ]
240
-
241
- # --- 右图: Baseline + 6个输出扰动 ---
242
- op_cfgs = [
243
- ("Baseline", "baseline", '#F04438'),
244
- ("OP(s=0.005)", "perturbation_0.005", '#A6F4C5'),
245
- ("OP(s=0.01)", "perturbation_0.01", '#6CE9A6'),
246
- ("OP(s=0.015)", "perturbation_0.015", '#32D583'),
247
- ("OP(s=0.02)", "perturbation_0.02", '#12B76A'),
248
- ("OP(s=0.025)", "perturbation_0.025", '#039855'),
249
- ("OP(s=0.03)", "perturbation_0.03", '#027A48'),
250
- ]
251
-
252
- for ax_idx, (ax, cfgs, title) in enumerate([
253
- (axes[0], ls_cfgs, 'Label Smoothing (5 models)'),
254
- (axes[1], op_cfgs, 'Output Perturbation (7 configs)')
255
- ]):
256
- ax.set_facecolor('white')
257
 
258
- # 计算归一化用的最大值(基于当前子图的配置)
259
- mx = []
260
- for i, m_key in enumerate(mk):
261
- val_max = max(gm(k, m_key) for _, k, _ in cfgs)
262
- mx.append(val_max if val_max > 0 else 1)
263
 
 
 
 
264
  for nm, ky, cl in cfgs:
265
- v = [gm(ky, m_key) / mx[i] for i, m_key in enumerate(mk)]
266
- v += [v[0]] # 闭合
267
- lw = 2.8 if ky == 'baseline' else 1.8
268
- alpha_fill = 0.10 if ky == 'baseline' else 0.04
269
- ax.plot(ag, v, 'o-', lw=lw, label=nm, color=cl, ms=5,
270
- alpha=0.95 if ky == 'baseline' else 0.85)
271
- ax.fill(ag, v, alpha=alpha_fill, color=cl)
272
-
273
- ax.set_xticks(ag[:-1])
274
- ax.set_xticklabels(ms, fontsize=9, color=COLORS['text'])
275
- ax.set_yticklabels([])
276
- ax.set_title(title, fontsize=11, fontweight='700',
277
- color=COLORS['text'], pad=18)
278
- ax.legend(loc='upper right',
279
- bbox_to_anchor=(1.35 if ax_idx == 1 else 1.30, 1.12),
280
- fontsize=8, framealpha=0.9, edgecolor=COLORS['grid'])
281
- ax.spines['polar'].set_color(COLORS['grid'])
282
- ax.grid(color=COLORS['grid'], alpha=0.5)
283
-
284
  plt.tight_layout()
285
  return fig
286
 
287
  def fig_loss_dist():
288
- items = [(k, l, gm(k, 'auc')) for k, l in zip(LS_KEYS, LS_LABELS_EN) if k in full_losses]; n = len(items)
289
  if n == 0: return plt.figure()
290
  fig, axes = plt.subplots(1, n, figsize=(4.5*n, 4.5)); axes = [axes] if n == 1 else axes; apply_light_style(fig, axes)
291
  for ax, (k, l, a) in zip(axes, items):
@@ -307,14 +239,14 @@ def fig_perturb_dist():
307
  ax.hist(mp, bins=bins, alpha=0.6, color=COLORS['accent'], label='Mem+noise', density=True, edgecolor='white')
308
  ax.hist(np_, bins=bins, alpha=0.6, color=COLORS['danger'], label='Non+noise', density=True, edgecolor='white')
309
  pa = gm(f'perturbation_{s}', 'auc')
310
- ax.set_title(f'OP(s={s})\nAUC={pa:.4f}', fontsize=11, fontweight='semibold'); ax.set_xlabel('Loss', fontsize=10)
311
  ax.legend(fontsize=9, facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text'])
312
  plt.tight_layout(); return fig
313
 
314
  def fig_roc_curves():
315
  fig, axes = plt.subplots(1, 2, figsize=(16, 7)); apply_light_style(fig, axes)
316
  ax = axes[0]; ls_colors = [COLORS['danger'], COLORS['ls_colors'][0], COLORS['ls_colors'][1], COLORS['ls_colors'][2], COLORS['ls_colors'][3]]
317
- for i, (k, l) in enumerate(zip(LS_KEYS, LS_LABELS_EN)):
318
  if k not in full_losses: continue
319
  m = np.array(full_losses[k]['member_losses']); nm = np.array(full_losses[k]['non_member_losses'])
320
  y_true = np.concatenate([np.ones(len(m)), np.zeros(len(nm))]); y_scores = np.concatenate([-m, -nm])
@@ -327,56 +259,58 @@ def fig_roc_curves():
327
  fpr, tpr, _ = roc_curve(y_true, y_scores); ax.plot(fpr, tpr, color=COLORS['danger'], lw=2.5, label=f'Baseline (AUC={bl_auc:.4f})')
328
  for i, s in enumerate(OP_SIGMAS):
329
  rng_m = np.random.RandomState(42); rng_nm = np.random.RandomState(137); mp = ml_base + rng_m.normal(0, s, len(ml_base)); np_ = nl_base + rng_nm.normal(0, s, len(nl_base)); y_scores_p = np.concatenate([-mp, -np_]); fpr_p, tpr_p, _ = roc_curve(y_true, y_scores_p); auc_p = roc_auc_score(y_true, y_scores_p)
330
- ax.plot(fpr_p, tpr_p, color=COLORS['op_colors'][i], lw=2, label=f'OP(s={s}) (AUC={auc_p:.4f})')
331
  ax.plot([0,1], [0,1], '--', color=COLORS['text_dim'], lw=1.5, label='Random'); ax.set_xlabel('False Positive Rate', fontsize=12, fontweight='medium'); ax.set_ylabel('True Positive Rate', fontsize=12, fontweight='medium'); ax.set_title('ROC Curves: Output Perturbation', fontsize=14, fontweight='bold', pad=15); ax.legend(fontsize=10, facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text'], loc='lower right'); plt.tight_layout()
332
  return fig
333
 
334
  def fig_tpr_at_low_fpr():
335
  fig, axes = plt.subplots(1, 2, figsize=(16, 6.5)); apply_light_style(fig, axes); labels_all, tpr5_all, tpr1_all, colors_all = [], [], [], []; ls_c = [COLORS['baseline']] + COLORS['ls_colors']
336
- for i, (k, l) in enumerate(zip(LS_KEYS, LS_LABELS_EN)): labels_all.append(l); tpr5_all.append(gm(k, 'tpr_at_5fpr')); tpr1_all.append(gm(k, 'tpr_at_1fpr')); colors_all.append(ls_c[i])
337
- for i, (k, l) in enumerate(zip(OP_KEYS, OP_LABELS_EN)): labels_all.append(l); tpr5_all.append(gm(k, 'tpr_at_5fpr')); tpr1_all.append(gm(k, 'tpr_at_1fpr')); colors_all.append(COLORS['op_colors'][i])
338
  x = range(len(labels_all)); ax = axes[0]; bars = ax.bar(x, tpr5_all, color=colors_all, width=0.65, edgecolor='none', zorder=3)
339
  for b, v in zip(bars, tpr5_all): ax.text(b.get_x()+b.get_width()/2, v+0.005, f'{v:.3f}', ha='center', fontsize=9, fontweight='semibold', color=COLORS['text'])
340
- ax.set_ylabel('TPR @ 5% FPR', fontsize=12, fontweight='medium'); ax.set_title('Attack Power at 5% FPR', fontsize=14, fontweight='bold', pad=15); ax.set_xticks(x); ax.set_xticklabels(labels_all, rotation=35, ha='right', fontsize=10); ax.axhline(0.05, color=COLORS['warning'], ls='--', lw=1.5, alpha=0.7, label='Random (0.05)'); ax.legend(facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text'], fontsize=10)
341
  ax = axes[1]; bars = ax.bar(x, tpr1_all, color=colors_all, width=0.65, edgecolor='none', zorder=3)
342
  for b, v in zip(bars, tpr1_all): ax.text(b.get_x()+b.get_width()/2, v+0.003, f'{v:.3f}', ha='center', fontsize=9, fontweight='semibold', color=COLORS['text'])
343
- ax.set_ylabel('TPR @ 1% FPR', fontsize=12, fontweight='medium'); ax.set_title('Attack Power at 1% FPR (Strict)', fontsize=14, fontweight='bold', pad=15); ax.set_xticks(x); ax.set_xticklabels(labels_all, rotation=35, ha='right', fontsize=10); ax.axhline(0.01, color=COLORS['warning'], ls='--', lw=1.5, alpha=0.7, label='Random (0.01)'); ax.legend(facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text'], fontsize=10); plt.tight_layout()
344
  return fig
345
 
346
  def fig_acc_bar():
347
  names, vals, clrs = [], [], []; ls_c = [COLORS['baseline']] + COLORS['ls_colors']
348
- for i, (k, l) in enumerate(zip(LS_KEYS, LS_LABELS_EN)):
349
  if k in utility_results: names.append(l); vals.append(utility_results[k]['accuracy']*100); clrs.append(ls_c[i])
350
- for i, (k, l) in enumerate(zip(OP_KEYS, OP_LABELS_EN)):
351
  if k in perturb_results: names.append(l); vals.append(bl_acc); clrs.append(COLORS['op_colors'][i])
352
  fig, ax = plt.subplots(figsize=(14, 6)); apply_light_style(fig, ax); bars = ax.bar(range(len(names)), vals, color=clrs, width=0.65, edgecolor='none', zorder=3)
353
  for b, v in zip(bars, vals): ax.text(b.get_x()+b.get_width()/2, v+1, f'{v:.1f}%', ha='center', fontsize=10, fontweight='semibold', color=COLORS['text'])
354
- ax.set_ylabel('Test Accuracy (%)', fontsize=12, fontweight='medium'); ax.set_title('Model Utility: Test Accuracy', fontsize=14, fontweight='bold', pad=20); ax.set_ylim(0, 105); ax.set_xticks(range(len(names))); ax.set_xticklabels(names, rotation=30, ha='right', fontsize=10); plt.tight_layout()
355
  return fig
356
 
357
  def fig_tradeoff():
358
  fig, ax = plt.subplots(figsize=(11, 8)); apply_light_style(fig, ax); markers_ls = ['o', 's', 's', 's', 's']; ls_c = [COLORS['baseline']] + COLORS['ls_colors']
359
- for i, (k, l) in enumerate(zip(LS_KEYS, LS_LABELS_EN)):
360
  if k in mia_results and k in utility_results: ax.scatter(utility_results[k]['accuracy']*100, mia_results[k]['auc'], label=l, marker=markers_ls[i], color=ls_c[i], s=180, edgecolors='white', lw=1.5, zorder=5, alpha=0.9)
361
  op_markers = ['^', 'D', 'v', 'P', 'X', 'h']
362
- for i, (k, l) in enumerate(zip(OP_KEYS, OP_LABELS_EN)):
363
  if k in perturb_results: ax.scatter(bl_acc, perturb_results[k]['auc'], label=l, marker=op_markers[i], color=COLORS['op_colors'][i], s=180, edgecolors='white', lw=1.5, zorder=5, alpha=0.9)
364
  ax.axhline(0.5, color=COLORS['text_dim'], ls='--', alpha=0.6, label='Random (AUC=0.5)'); ax.annotate('IDEAL ZONE\nHigh Utility, Low Risk', xy=(85, 0.51), fontsize=11, fontweight='bold', color=COLORS['success'], alpha=0.7, ha='center', backgroundcolor=COLORS['bg']); ax.annotate('HIGH RISK ZONE\nLow Utility, High Risk', xy=(62, 0.61), fontsize=11, fontweight='bold', color=COLORS['danger'], alpha=0.7, ha='center', backgroundcolor=COLORS['bg']); ax.set_xlabel('Model Utility (Accuracy %)', fontsize=12, fontweight='medium'); ax.set_ylabel('Privacy Risk (MIA AUC)', fontsize=12, fontweight='medium'); ax.set_title('Privacy-Utility Trade-off Analysis', fontsize=14, fontweight='bold', pad=20); ax.legend(fontsize=10, loc='upper left', ncol=2, facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text']); plt.tight_layout()
365
  return fig
366
 
367
  def fig_auc_trend():
368
  fig, axes = plt.subplots(1, 2, figsize=(16, 6.5)); apply_light_style(fig, axes); ax = axes[0]; eps_vals = [0.0, 0.02, 0.05, 0.1, 0.2]; auc_vals = [gm(k, 'auc') for k in LS_KEYS]; acc_vals = [gu(k) for k in LS_KEYS]
369
- ax2 = ax.twinx(); line1 = ax.plot(eps_vals, auc_vals, 'o-', color=COLORS['danger'], lw=3, ms=9, label='MIA AUC (left)', zorder=5); line2 = ax2.plot(eps_vals, acc_vals, 's--', color=COLORS['accent'], lw=3, ms=9, label='Utility % (right)', zorder=5); ax.axhline(0.5, color=COLORS['text_dim'], ls=':', alpha=0.5); ax.set_xlabel('Label Smoothing epsilon', fontsize=12, fontweight='medium'); ax.set_ylabel('MIA AUC', fontsize=12, fontweight='medium', color=COLORS['danger']); ax2.set_ylabel('Utility (%)', fontsize=12, fontweight='medium', color=COLORS['accent']); ax.set_title('Label Smoothing Trends', fontsize=14, fontweight='bold', pad=15); ax.tick_params(axis='y', labelcolor=COLORS['danger']); ax2.tick_params(axis='y', labelcolor=COLORS['accent']); ax2.spines['right'].set_color(COLORS['accent']); ax2.spines['left'].set_color(COLORS['danger']); lines = line1 + line2; labels = [l.get_label() for l in lines]; ax.legend(lines, labels, fontsize=10, facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text'])
370
- ax = axes[1]; sig_vals = OP_SIGMAS; auc_op = [gm(k, 'auc') for k in OP_KEYS]; ax.plot(sig_vals, auc_op, 'o-', color=COLORS['success'], lw=3, ms=9, zorder=5, label='MIA AUC'); ax.axhline(bl_auc, color=COLORS['danger'], ls='--', lw=2, alpha=0.6, label=f'Baseline ({bl_auc:.4f})'); ax.axhline(0.5, color=COLORS['text_dim'], ls=':', alpha=0.5, label='Random (0.5)'); ax.fill_between(sig_vals, auc_op, bl_auc, alpha=0.2, color=COLORS['success'], label='AUC Reduction'); ax.set_xlabel('Perturbation Sigma', fontsize=12, fontweight='medium'); ax.set_ylabel('MIA AUC', fontsize=12, fontweight='medium'); ax.set_title('Output Perturbation Trends', fontsize=14, fontweight='bold', pad=15); ax.legend(fontsize=10, facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text']); plt.tight_layout()
 
 
371
  return fig
372
 
373
  def fig_loss_gap_waterfall():
374
  fig, ax = plt.subplots(figsize=(14, 6.5)); apply_light_style(fig, ax); names, gaps, clrs = [], [], []; ls_c = [COLORS['baseline']] + COLORS['ls_colors']
375
- for i, (k, l) in enumerate(zip(LS_KEYS, LS_LABELS_EN)): names.append(l); gaps.append(gm(k, 'loss_gap')); clrs.append(ls_c[i])
376
- for i, (k, l) in enumerate(zip(OP_KEYS, OP_LABELS_EN)): names.append(l); gaps.append(gm(k, 'loss_gap')); clrs.append(COLORS['op_colors'][i])
377
  bars = ax.bar(range(len(names)), gaps, color=clrs, width=0.65, edgecolor='none', zorder=3)
378
  for b, v in zip(bars, gaps): ax.text(b.get_x()+b.get_width()/2, v+0.0005, f'{v:.4f}', ha='center', fontsize=10, fontweight='semibold', color=COLORS['text'])
379
- ax.set_ylabel('Loss Gap', fontsize=12, fontweight='medium'); ax.set_title('Member vs Non-Member Loss Gap', fontsize=14, fontweight='bold', pad=20); ax.set_xticks(range(len(names))); ax.set_xticklabels(names, rotation=30, ha='right', fontsize=10); ax.annotate('Smaller gap = Better Privacy', xy=(8, gaps[0]*0.4), fontsize=11, color=COLORS['success'], fontstyle='italic', ha='center', backgroundcolor=COLORS['bg'], bbox=dict(boxstyle='round,pad=0.4', facecolor=COLORS['panel'], edgecolor=COLORS['success'], alpha=0.8)); plt.tight_layout()
380
  return fig
381
 
382
  # ================================================================
@@ -401,6 +335,7 @@ def cb_sample(src):
401
  """
402
  return md, clean_text(s.get('question', '')), clean_text(s.get('answer', ''))
403
 
 
404
  ATK_CHOICES = (
405
  ["基线模型 (Baseline)"] +
406
  [f"标签平滑 (ε={e})" for e in [0.02, 0.05, 0.1, 0.2]] +
@@ -517,12 +452,12 @@ def cb_eval(model_choice):
517
 
518
  def build_full_table():
519
  rows = []
520
- for k, l in zip(LS_KEYS, LS_LABELS_CN):
521
  if k in mia_results:
522
  m = mia_results[k]; u = gu(k)
523
  t = "—" if k == "baseline" else "训练期"; d = "" if k == "baseline" else f"{m['auc']-bl_auc:+.4f}"
524
  rows.append(f"| {l} | {t} | {m['auc']:.4f} | {m['attack_accuracy']:.4f} | {m['precision']:.4f} | {m['recall']:.4f} | {m['f1']:.4f} | {m['tpr_at_5fpr']:.4f} | {m['tpr_at_1fpr']:.4f} | {m['loss_gap']:.4f} | {u:.1f}% | {d} |")
525
- for k, l in zip(OP_KEYS, OP_LABELS_CN):
526
  if k in perturb_results:
527
  m = perturb_results[k]; d = f"{m['auc']-bl_auc:+.4f}"
528
  rows.append(f"| {l} | 推理期 | {m['auc']:.4f} | {m['attack_accuracy']:.4f} | {m['precision']:.4f} | {m['recall']:.4f} | {m['f1']:.4f} | {m['tpr_at_5fpr']:.4f} | {m['tpr_at_1fpr']:.4f} | {m['loss_gap']:.4f} | {bl_acc:.1f}% | {d} |")
@@ -605,8 +540,8 @@ with gr.Blocks(title="MIA攻防研究") as demo:
605
  """)
606
 
607
  # 实验体系总览图 (如果在目录里则显示)
608
- if os.path.exists(os.path.join(BASE_DIR, "figures", "algo4_overview.png")):
609
- gr.Image(os.path.join(BASE_DIR, "figures", "algo4_overview.png"), label="实验体系总览", show_label=True)
610
 
611
  gr.HTML(f"""<div style="display:grid;grid-template-columns:repeat(4,1fr);gap:20px;margin:30px 0;">
612
  <div class="card-wrap" style="text-align:center;">
@@ -719,24 +654,22 @@ with gr.Blocks(title="MIA攻防研究") as demo:
719
  gr.Markdown(f"### 1️⃣ 攻击成功率全景对比 (AUC)\n\n> 柱子越短 = AUC越低 = 防御越有效。基线AUC={bl_auc:.4f},标签平滑最低降至{gm('smooth_eps_0.2','auc'):.4f},输出扰动最低降至{gm('perturbation_0.03','auc'):.4f}。")
720
  gr.Plot(value=fig_auc_bar())
721
 
722
- # --- 整合双雷达图及配套讲解文本 ---
723
  gr.Markdown(f"""\
724
  ### 2️⃣ 多指标雷达图对比(全部11组实验)
725
 
726
  > **左图:标签平滑系列5个模型**
727
  > - 红色(Baseline)面积最大 = 攻击全面有效
728
- > - 随着ε从0.020.2增大,雷达面积逐步缩小 = 防御逐步增强
729
  > - 特别注意 TPR@1%FPR 和 LossGap 两个轴,缩小最显著
730
  >
731
  > **右图:输出扰动系列7个配置**
732
  > - 红色(Baseline)同样是最大的
733
- > - 随着σ从0.0050.03增大,绿色系雷达逐步缩小
734
  > - OP在LossGap和TPR@5%维度上降幅尤其明显
735
  >
736
  > **结论:** 两种防御均在所有维度上全面压制攻击能力,不是只降低了某一个指标。
737
  """)
738
  gr.Plot(value=fig_radar())
739
- # ---------------------------------
740
 
741
  gr.Markdown("### 3️⃣ ROC曲线对比\n\n> 曲线越贴近对角线=攻击越接近随机猜测=防御越有效。左图标签平滑,右图输出扰动。")
742
  gr.Plot(value=fig_roc_curves())
@@ -862,4 +795,4 @@ with gr.Blocks(title="MIA攻防研究") as demo:
862
 
863
  """)
864
 
865
- demo.launch(theme=gr.themes.Soft(), css=CSS)
 
1
  # ================================================================
2
+ # 教育大模型MIA攻防研究 - Gradio演示系统 v7.0 学术巅峰版
3
+ # 彻底消灭普通 e/s,全量启用 LaTeX 原生学斜体 $\epsilon$ 和 $\sigma$
4
  # ================================================================
5
 
6
  import os
 
31
  text = re.sub(r'[\u200b-\u200f\u2028-\u202f\u2060-\u206f\ufeff]', '', text)
32
  return text.strip()
33
 
 
34
  try:
35
  member_data = load_json("data/member.json")
36
  non_member_data = load_json("data/non_member.json")
 
57
  for s in [0.005, 0.01, 0.015, 0.02, 0.025, 0.03]:
58
  k = f"perturbation_{s}"
59
  perturb_results[k] = {m: v*0.85 for m, v in mia_results["baseline"].items()}
 
60
  perturb_results[k]["member_loss_std"] = np.sqrt(0.03**2 + s**2)
61
  perturb_results[k]["non_member_loss_std"] = np.sqrt(0.03**2 + s**2)
62
 
63
  # ================================================================
64
+ # 全局图表配置
65
  # ================================================================
66
  COLORS = {
67
  'bg': '#FFFFFF',
 
78
  'ls_colors': ['#A0C4FF', '#70A1FF', '#478EFF', '#007AFF'],
79
  'op_colors': ['#98F5E1', '#6EE7B7', '#34D399', '#10B981', '#059669', '#047857'],
80
  }
 
 
81
  CHART_W = 14
82
 
83
  def apply_light_style(fig, ax_or_axes):
 
96
  ax.grid(True, color=COLORS['grid'], alpha=0.6, linestyle='-', linewidth=0.8)
97
  ax.set_axisbelow(True)
98
 
99
+ # 🌟🌟🌟 核心修改:专门为画图准备的 LaTeX 格式标签 🌟🌟🌟
 
 
100
  LS_KEYS = ["baseline", "smooth_eps_0.02", "smooth_eps_0.05", "smooth_eps_0.1", "smooth_eps_0.2"]
101
+ LS_LABELS_PLOT = ["Baseline", r"LS($\epsilon$=0.02)", r"LS($\epsilon$=0.05)", r"LS($\epsilon$=0.1)", r"LS($\epsilon$=0.2)"]
102
+ LS_LABELS_UI = ["基线(Baseline)", "LS(ε=0.02)", "LS(ε=0.05)", "LS(ε=0.1)", "LS(ε=0.2)"]
103
 
104
  OP_SIGMAS = [0.005, 0.01, 0.015, 0.02, 0.025, 0.03]
105
  OP_KEYS = [f"perturbation_{s}" for s in OP_SIGMAS]
106
+ OP_LABELS_PLOT = [f"OP($\sigma$={s})" for s in OP_SIGMAS]
107
+ OP_LABELS_UI = [f"OP(σ={s})" for s in OP_SIGMAS]
108
 
109
  ALL_KEYS = LS_KEYS + OP_KEYS
110
 
 
123
  bl_m_mean = gm("baseline", "member_loss_mean")
124
  bl_nm_mean = gm("baseline", "non_member_loss_mean")
125
 
126
+ TYPE_CN = {'calculation': '基础计算', 'word_problem': '应用题', 'concept': '概念问答', 'error_correction': '错题订正'}
 
127
 
 
 
 
128
  np.random.seed(777)
129
  EVAL_POOL = []
130
  _types = ['calculation']*120 + ['word_problem']*90 + ['concept']*60 + ['error_correction']*30
 
138
  else: _q,_ans=f"{_a} x {_b} = ?",str(_a*_b)
139
  elif _t == 'word_problem':
140
  _a,_b = int(np.random.randint(5,200)), int(np.random.randint(3,50))
141
+ _tpls = [(f"{_a} apples, ate {_b}, left?",str(_a-_b)), (f"{_a} per group, {_b} groups, total?",str(_a*_b))]
 
142
  _q,_ans = _tpls[_i%len(_tpls)]
143
  elif _t == 'concept':
144
  _cs = [("area","Area = space occupied by a shape"),("perimeter","Perimeter = total boundary length")]
 
153
  EVAL_POOL.append(item)
154
 
155
  # ================================================================
156
+ # 图表绘制函数 (全部更换为 LaTeX 渲染)
157
  # ================================================================
158
  def fig_gauge(loss_val, m_mean, nm_mean, thr, m_std, nm_std):
159
+ fig, ax = plt.subplots(figsize=(10, 2.6)); fig.patch.set_facecolor(COLORS['bg']); ax.set_facecolor(COLORS['panel'])
160
+ xlo = min(m_mean - 3.0 * m_std, loss_val - 0.005); xhi = max(nm_mean + 3.0 * nm_std, loss_val + 0.005)
161
+ ax.axvspan(xlo, thr, alpha=0.2, color=COLORS['accent']); ax.axvspan(thr, xhi, alpha=0.2, color=COLORS['danger'])
 
 
 
 
 
 
 
162
  ax.axvline(m_mean, color=COLORS['accent'], lw=2, ls=':', alpha=0.8, zorder=2)
163
  ax.text(m_mean - 0.002, 1.02, f'Member Mean\n{m_mean:.4f}', ha='right', va='bottom', fontsize=9, color=COLORS['accent'], transform=ax.get_xaxis_transform())
 
164
  ax.axvline(nm_mean, color=COLORS['danger'], lw=2, ls=':', alpha=0.8, zorder=2)
165
  ax.text(nm_mean + 0.002, 1.02, f'Non-Member Mean\n{nm_mean:.4f}', ha='left', va='bottom', fontsize=9, color=COLORS['danger'], transform=ax.get_xaxis_transform())
 
166
  ax.axvline(thr, color=COLORS['text_dim'], lw=2.5, ls='--', zorder=3)
167
  ax.text(thr, 1.25, f'Threshold\n{thr:.4f}', ha='center', va='bottom', fontsize=10, fontweight='bold', color=COLORS['text_dim'], transform=ax.get_xaxis_transform())
 
168
  mc = COLORS['accent'] if loss_val < thr else COLORS['danger']
169
  ax.plot(loss_val, 0.5, marker='o', ms=16, color='white', mec=mc, mew=3, zorder=5, transform=ax.get_xaxis_transform())
170
  ax.text(loss_val, 0.75, f'Current Loss\n{loss_val:.4f}', ha='center', fontsize=11, fontweight='bold', color=mc, transform=ax.get_xaxis_transform())
 
171
  ax.text((xlo+thr)/2, 0.25, 'MEMBER', ha='center', fontsize=12, color=COLORS['accent'], alpha=0.6, fontweight='bold', transform=ax.get_xaxis_transform())
172
  ax.text((thr+xhi)/2, 0.25, 'NON-MEMBER', ha='center', fontsize=12, color=COLORS['danger'], alpha=0.6, fontweight='bold', transform=ax.get_xaxis_transform())
 
173
  ax.set_xlim(xlo, xhi); ax.set_yticks([])
174
  for s in ax.spines.values(): s.set_visible(False)
175
+ ax.spines['bottom'].set_visible(True); ax.spines['bottom'].set_color(COLORS['grid']); ax.tick_params(colors=COLORS['text_dim'], width=1)
176
+ ax.set_xlabel('Loss Value', fontsize=11, color=COLORS['text'], fontweight='medium'); plt.tight_layout(pad=0.5)
 
 
177
  return fig
178
 
179
  def fig_auc_bar():
180
  names, vals, clrs = [], [], []
181
  ls_c = [COLORS['baseline']] + COLORS['ls_colors']
182
+ for i,(k,l) in enumerate(zip(LS_KEYS, LS_LABELS_PLOT)):
183
  if k in mia_results: names.append(l); vals.append(mia_results[k]['auc']); clrs.append(ls_c[i])
184
+ for i,(k,l) in enumerate(zip(OP_KEYS, OP_LABELS_PLOT)):
185
  if k in perturb_results: names.append(l); vals.append(perturb_results[k]['auc']); clrs.append(COLORS['op_colors'][i])
186
  fig, ax = plt.subplots(figsize=(14, 6)); apply_light_style(fig, ax)
187
  bars = ax.bar(range(len(names)), vals, color=clrs, width=0.65, edgecolor='none', zorder=3)
 
189
  ax.axhline(0.5, color=COLORS['text_dim'], ls='--', lw=1.5, alpha=0.6, label='Random Guess (0.5)', zorder=2)
190
  ax.axhline(bl_auc, color=COLORS['danger'], ls=':', lw=1.5, alpha=0.8, label=f'Baseline ({bl_auc:.4f})', zorder=2)
191
  ax.set_ylabel('MIA Attack AUC', fontsize=12, fontweight='medium'); ax.set_title('Defense Effectiveness: MIA AUC Comparison', fontsize=14, fontweight='bold', pad=20)
192
+ ax.set_ylim(0.45, max(vals)+0.05); ax.set_xticks(range(len(names))); ax.set_xticklabels(names, rotation=30, ha='right', fontsize=11)
193
  ax.legend(facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text'], fontsize=10, loc='upper right'); plt.tight_layout()
194
  return fig
195
 
196
  def fig_radar():
197
  ms = ['AUC', 'Atk Acc', 'Prec', 'Recall', 'F1', 'TPR@5%', 'TPR@1%', 'Gap']
198
+ mk = ['auc', 'attack_accuracy', 'precision', 'recall', 'f1', 'tpr_at_5fpr', 'tpr_at_1fpr', 'loss_gap']
199
+ N = len(ms); ag = np.linspace(0, 2 * np.pi, N, endpoint=False).tolist() + [0]
200
+ fig, axes = plt.subplots(1, 2, figsize=(CHART_W + 2, 7), subplot_kw=dict(polar=True)); fig.patch.set_facecolor('white')
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
201
 
202
+ ls_cfgs = [("Baseline", "baseline", '#F04438'), (r"LS($\epsilon$=0.02)", "smooth_eps_0.02", '#B2DDFF'), (r"LS($\epsilon$=0.05)", "smooth_eps_0.05", '#84CAFF'), (r"LS($\epsilon$=0.1)", "smooth_eps_0.1", '#2E90FA'), (r"LS($\epsilon$=0.2)", "smooth_eps_0.2", '#7A5AF8')]
203
+ op_cfgs = [("Baseline", "baseline", '#F04438'), (r"OP($\sigma$=0.005)", "perturbation_0.005", '#A6F4C5'), (r"OP($\sigma$=0.01)", "perturbation_0.01", '#6CE9A6'), (r"OP($\sigma$=0.015)", "perturbation_0.015", '#32D583'), (r"OP($\sigma$=0.02)", "perturbation_0.02", '#12B76A'), (r"OP($\sigma$=0.025)", "perturbation_0.025", '#039855'), (r"OP($\sigma$=0.03)", "perturbation_0.03", '#027A48')]
 
 
 
204
 
205
+ for ax_idx, (ax, cfgs, title) in enumerate([(axes[0], ls_cfgs, 'Label Smoothing (5 models)'), (axes[1], op_cfgs, 'Output Perturbation (7 configs)')]):
206
+ ax.set_facecolor('white')
207
+ mx = [max(gm(k, m_key) for _, k, _ in cfgs) for m_key in mk]; mx = [m if m > 0 else 1 for m in mx]
208
  for nm, ky, cl in cfgs:
209
+ v = [gm(ky, m_key) / mx[i] for i, m_key in enumerate(mk)]; v += [v[0]]
210
+ ax.plot(ag, v, 'o-', lw=2.8 if ky == 'baseline' else 1.8, label=nm, color=cl, ms=5, alpha=0.95 if ky == 'baseline' else 0.85)
211
+ ax.fill(ag, v, alpha=0.10 if ky == 'baseline' else 0.04, color=cl)
212
+ ax.set_xticks(ag[:-1]); ax.set_xticklabels(ms, fontsize=10, color=COLORS['text']); ax.set_yticklabels([])
213
+ ax.set_title(title, fontsize=12, fontweight='700', color=COLORS['text'], pad=18)
214
+ ax.legend(loc='upper right', bbox_to_anchor=(1.35 if ax_idx == 1 else 1.30, 1.12), fontsize=9, framealpha=0.9, edgecolor=COLORS['grid'])
215
+ ax.spines['polar'].set_color(COLORS['grid']); ax.grid(color=COLORS['grid'], alpha=0.5)
 
 
 
 
 
 
 
 
 
 
 
 
216
  plt.tight_layout()
217
  return fig
218
 
219
  def fig_loss_dist():
220
+ items = [(k, l, gm(k, 'auc')) for k, l in zip(LS_KEYS, LS_LABELS_PLOT) if k in full_losses]; n = len(items)
221
  if n == 0: return plt.figure()
222
  fig, axes = plt.subplots(1, n, figsize=(4.5*n, 4.5)); axes = [axes] if n == 1 else axes; apply_light_style(fig, axes)
223
  for ax, (k, l, a) in zip(axes, items):
 
239
  ax.hist(mp, bins=bins, alpha=0.6, color=COLORS['accent'], label='Mem+noise', density=True, edgecolor='white')
240
  ax.hist(np_, bins=bins, alpha=0.6, color=COLORS['danger'], label='Non+noise', density=True, edgecolor='white')
241
  pa = gm(f'perturbation_{s}', 'auc')
242
+ ax.set_title(f'OP($\sigma$={s})\nAUC={pa:.4f}', fontsize=11, fontweight='semibold'); ax.set_xlabel('Loss', fontsize=10)
243
  ax.legend(fontsize=9, facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text'])
244
  plt.tight_layout(); return fig
245
 
246
  def fig_roc_curves():
247
  fig, axes = plt.subplots(1, 2, figsize=(16, 7)); apply_light_style(fig, axes)
248
  ax = axes[0]; ls_colors = [COLORS['danger'], COLORS['ls_colors'][0], COLORS['ls_colors'][1], COLORS['ls_colors'][2], COLORS['ls_colors'][3]]
249
+ for i, (k, l) in enumerate(zip(LS_KEYS, LS_LABELS_PLOT)):
250
  if k not in full_losses: continue
251
  m = np.array(full_losses[k]['member_losses']); nm = np.array(full_losses[k]['non_member_losses'])
252
  y_true = np.concatenate([np.ones(len(m)), np.zeros(len(nm))]); y_scores = np.concatenate([-m, -nm])
 
259
  fpr, tpr, _ = roc_curve(y_true, y_scores); ax.plot(fpr, tpr, color=COLORS['danger'], lw=2.5, label=f'Baseline (AUC={bl_auc:.4f})')
260
  for i, s in enumerate(OP_SIGMAS):
261
  rng_m = np.random.RandomState(42); rng_nm = np.random.RandomState(137); mp = ml_base + rng_m.normal(0, s, len(ml_base)); np_ = nl_base + rng_nm.normal(0, s, len(nl_base)); y_scores_p = np.concatenate([-mp, -np_]); fpr_p, tpr_p, _ = roc_curve(y_true, y_scores_p); auc_p = roc_auc_score(y_true, y_scores_p)
262
+ ax.plot(fpr_p, tpr_p, color=COLORS['op_colors'][i], lw=2, label=f'OP($\sigma$={s}) (AUC={auc_p:.4f})')
263
  ax.plot([0,1], [0,1], '--', color=COLORS['text_dim'], lw=1.5, label='Random'); ax.set_xlabel('False Positive Rate', fontsize=12, fontweight='medium'); ax.set_ylabel('True Positive Rate', fontsize=12, fontweight='medium'); ax.set_title('ROC Curves: Output Perturbation', fontsize=14, fontweight='bold', pad=15); ax.legend(fontsize=10, facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text'], loc='lower right'); plt.tight_layout()
264
  return fig
265
 
266
  def fig_tpr_at_low_fpr():
267
  fig, axes = plt.subplots(1, 2, figsize=(16, 6.5)); apply_light_style(fig, axes); labels_all, tpr5_all, tpr1_all, colors_all = [], [], [], []; ls_c = [COLORS['baseline']] + COLORS['ls_colors']
268
+ for i, (k, l) in enumerate(zip(LS_KEYS, LS_LABELS_PLOT)): labels_all.append(l); tpr5_all.append(gm(k, 'tpr_at_5fpr')); tpr1_all.append(gm(k, 'tpr_at_1fpr')); colors_all.append(ls_c[i])
269
+ for i, (k, l) in enumerate(zip(OP_KEYS, OP_LABELS_PLOT)): labels_all.append(l); tpr5_all.append(gm(k, 'tpr_at_5fpr')); tpr1_all.append(gm(k, 'tpr_at_1fpr')); colors_all.append(COLORS['op_colors'][i])
270
  x = range(len(labels_all)); ax = axes[0]; bars = ax.bar(x, tpr5_all, color=colors_all, width=0.65, edgecolor='none', zorder=3)
271
  for b, v in zip(bars, tpr5_all): ax.text(b.get_x()+b.get_width()/2, v+0.005, f'{v:.3f}', ha='center', fontsize=9, fontweight='semibold', color=COLORS['text'])
272
+ ax.set_ylabel('TPR @ 5% FPR', fontsize=12, fontweight='medium'); ax.set_title('Attack Power at 5% FPR', fontsize=14, fontweight='bold', pad=15); ax.set_xticks(x); ax.set_xticklabels(labels_all, rotation=35, ha='right', fontsize=11); ax.axhline(0.05, color=COLORS['warning'], ls='--', lw=1.5, alpha=0.7, label='Random (0.05)'); ax.legend(facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text'], fontsize=10)
273
  ax = axes[1]; bars = ax.bar(x, tpr1_all, color=colors_all, width=0.65, edgecolor='none', zorder=3)
274
  for b, v in zip(bars, tpr1_all): ax.text(b.get_x()+b.get_width()/2, v+0.003, f'{v:.3f}', ha='center', fontsize=9, fontweight='semibold', color=COLORS['text'])
275
+ ax.set_ylabel('TPR @ 1% FPR', fontsize=12, fontweight='medium'); ax.set_title('Attack Power at 1% FPR (Strict)', fontsize=14, fontweight='bold', pad=15); ax.set_xticks(x); ax.set_xticklabels(labels_all, rotation=35, ha='right', fontsize=11); ax.axhline(0.01, color=COLORS['warning'], ls='--', lw=1.5, alpha=0.7, label='Random (0.01)'); ax.legend(facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text'], fontsize=10); plt.tight_layout()
276
  return fig
277
 
278
  def fig_acc_bar():
279
  names, vals, clrs = [], [], []; ls_c = [COLORS['baseline']] + COLORS['ls_colors']
280
+ for i, (k, l) in enumerate(zip(LS_KEYS, LS_LABELS_PLOT)):
281
  if k in utility_results: names.append(l); vals.append(utility_results[k]['accuracy']*100); clrs.append(ls_c[i])
282
+ for i, (k, l) in enumerate(zip(OP_KEYS, OP_LABELS_PLOT)):
283
  if k in perturb_results: names.append(l); vals.append(bl_acc); clrs.append(COLORS['op_colors'][i])
284
  fig, ax = plt.subplots(figsize=(14, 6)); apply_light_style(fig, ax); bars = ax.bar(range(len(names)), vals, color=clrs, width=0.65, edgecolor='none', zorder=3)
285
  for b, v in zip(bars, vals): ax.text(b.get_x()+b.get_width()/2, v+1, f'{v:.1f}%', ha='center', fontsize=10, fontweight='semibold', color=COLORS['text'])
286
+ ax.set_ylabel('Test Accuracy (%)', fontsize=12, fontweight='medium'); ax.set_title('Model Utility: Test Accuracy', fontsize=14, fontweight='bold', pad=20); ax.set_ylim(0, 105); ax.set_xticks(range(len(names))); ax.set_xticklabels(names, rotation=30, ha='right', fontsize=11); plt.tight_layout()
287
  return fig
288
 
289
  def fig_tradeoff():
290
  fig, ax = plt.subplots(figsize=(11, 8)); apply_light_style(fig, ax); markers_ls = ['o', 's', 's', 's', 's']; ls_c = [COLORS['baseline']] + COLORS['ls_colors']
291
+ for i, (k, l) in enumerate(zip(LS_KEYS, LS_LABELS_PLOT)):
292
  if k in mia_results and k in utility_results: ax.scatter(utility_results[k]['accuracy']*100, mia_results[k]['auc'], label=l, marker=markers_ls[i], color=ls_c[i], s=180, edgecolors='white', lw=1.5, zorder=5, alpha=0.9)
293
  op_markers = ['^', 'D', 'v', 'P', 'X', 'h']
294
+ for i, (k, l) in enumerate(zip(OP_KEYS, OP_LABELS_PLOT)):
295
  if k in perturb_results: ax.scatter(bl_acc, perturb_results[k]['auc'], label=l, marker=op_markers[i], color=COLORS['op_colors'][i], s=180, edgecolors='white', lw=1.5, zorder=5, alpha=0.9)
296
  ax.axhline(0.5, color=COLORS['text_dim'], ls='--', alpha=0.6, label='Random (AUC=0.5)'); ax.annotate('IDEAL ZONE\nHigh Utility, Low Risk', xy=(85, 0.51), fontsize=11, fontweight='bold', color=COLORS['success'], alpha=0.7, ha='center', backgroundcolor=COLORS['bg']); ax.annotate('HIGH RISK ZONE\nLow Utility, High Risk', xy=(62, 0.61), fontsize=11, fontweight='bold', color=COLORS['danger'], alpha=0.7, ha='center', backgroundcolor=COLORS['bg']); ax.set_xlabel('Model Utility (Accuracy %)', fontsize=12, fontweight='medium'); ax.set_ylabel('Privacy Risk (MIA AUC)', fontsize=12, fontweight='medium'); ax.set_title('Privacy-Utility Trade-off Analysis', fontsize=14, fontweight='bold', pad=20); ax.legend(fontsize=10, loc='upper left', ncol=2, facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text']); plt.tight_layout()
297
  return fig
298
 
299
  def fig_auc_trend():
300
  fig, axes = plt.subplots(1, 2, figsize=(16, 6.5)); apply_light_style(fig, axes); ax = axes[0]; eps_vals = [0.0, 0.02, 0.05, 0.1, 0.2]; auc_vals = [gm(k, 'auc') for k in LS_KEYS]; acc_vals = [gu(k) for k in LS_KEYS]
301
+ ax2 = ax.twinx(); line1 = ax.plot(eps_vals, auc_vals, 'o-', color=COLORS['danger'], lw=3, ms=9, label='MIA AUC (left)', zorder=5); line2 = ax2.plot(eps_vals, acc_vals, 's--', color=COLORS['accent'], lw=3, ms=9, label='Utility % (right)', zorder=5); ax.axhline(0.5, color=COLORS['text_dim'], ls=':', alpha=0.5)
302
+ ax.set_xlabel(r'Label Smoothing $\epsilon$', fontsize=12, fontweight='medium'); ax.set_ylabel('MIA AUC', fontsize=12, fontweight='medium', color=COLORS['danger']); ax2.set_ylabel('Utility (%)', fontsize=12, fontweight='medium', color=COLORS['accent']); ax.set_title('Label Smoothing Trends', fontsize=14, fontweight='bold', pad=15); ax.tick_params(axis='y', labelcolor=COLORS['danger']); ax2.tick_params(axis='y', labelcolor=COLORS['accent']); ax2.spines['right'].set_color(COLORS['accent']); ax2.spines['left'].set_color(COLORS['danger']); lines = line1 + line2; labels = [l.get_label() for l in lines]; ax.legend(lines, labels, fontsize=10, facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text'])
303
+ ax = axes[1]; sig_vals = OP_SIGMAS; auc_op = [gm(k, 'auc') for k in OP_KEYS]; ax.plot(sig_vals, auc_op, 'o-', color=COLORS['success'], lw=3, ms=9, zorder=5, label='MIA AUC'); ax.axhline(bl_auc, color=COLORS['danger'], ls='--', lw=2, alpha=0.6, label=f'Baseline ({bl_auc:.4f})'); ax.axhline(0.5, color=COLORS['text_dim'], ls=':', alpha=0.5, label='Random (0.5)'); ax.fill_between(sig_vals, auc_op, bl_auc, alpha=0.2, color=COLORS['success'], label='AUC Reduction')
304
+ ax.set_xlabel(r'Perturbation $\sigma$', fontsize=12, fontweight='medium'); ax.set_ylabel('MIA AUC', fontsize=12, fontweight='medium'); ax.set_title('Output Perturbation Trends', fontsize=14, fontweight='bold', pad=15); ax.legend(fontsize=10, facecolor=COLORS['bg'], edgecolor='none', labelcolor=COLORS['text']); plt.tight_layout()
305
  return fig
306
 
307
  def fig_loss_gap_waterfall():
308
  fig, ax = plt.subplots(figsize=(14, 6.5)); apply_light_style(fig, ax); names, gaps, clrs = [], [], []; ls_c = [COLORS['baseline']] + COLORS['ls_colors']
309
+ for i, (k, l) in enumerate(zip(LS_KEYS, LS_LABELS_PLOT)): names.append(l); gaps.append(gm(k, 'loss_gap')); clrs.append(ls_c[i])
310
+ for i, (k, l) in enumerate(zip(OP_KEYS, OP_LABELS_PLOT)): names.append(l); gaps.append(gm(k, 'loss_gap')); clrs.append(COLORS['op_colors'][i])
311
  bars = ax.bar(range(len(names)), gaps, color=clrs, width=0.65, edgecolor='none', zorder=3)
312
  for b, v in zip(bars, gaps): ax.text(b.get_x()+b.get_width()/2, v+0.0005, f'{v:.4f}', ha='center', fontsize=10, fontweight='semibold', color=COLORS['text'])
313
+ ax.set_ylabel('Loss Gap', fontsize=12, fontweight='medium'); ax.set_title('Member vs Non-Member Loss Gap', fontsize=14, fontweight='bold', pad=20); ax.set_xticks(range(len(names))); ax.set_xticklabels(names, rotation=30, ha='right', fontsize=11); ax.annotate('Smaller gap = Better Privacy', xy=(8, gaps[0]*0.4), fontsize=11, color=COLORS['success'], fontstyle='italic', ha='center', backgroundcolor=COLORS['bg'], bbox=dict(boxstyle='round,pad=0.4', facecolor=COLORS['panel'], edgecolor=COLORS['success'], alpha=0.8)); plt.tight_layout()
314
  return fig
315
 
316
  # ================================================================
 
335
  """
336
  return md, clean_text(s.get('question', '')), clean_text(s.get('answer', ''))
337
 
338
+ # 🌟 下拉框选项也全部替换为 ε 和 σ
339
  ATK_CHOICES = (
340
  ["基线模型 (Baseline)"] +
341
  [f"标签平滑 (ε={e})" for e in [0.02, 0.05, 0.1, 0.2]] +
 
452
 
453
  def build_full_table():
454
  rows = []
455
+ for k, l in zip(LS_KEYS, LS_LABELS_UI):
456
  if k in mia_results:
457
  m = mia_results[k]; u = gu(k)
458
  t = "—" if k == "baseline" else "训练期"; d = "" if k == "baseline" else f"{m['auc']-bl_auc:+.4f}"
459
  rows.append(f"| {l} | {t} | {m['auc']:.4f} | {m['attack_accuracy']:.4f} | {m['precision']:.4f} | {m['recall']:.4f} | {m['f1']:.4f} | {m['tpr_at_5fpr']:.4f} | {m['tpr_at_1fpr']:.4f} | {m['loss_gap']:.4f} | {u:.1f}% | {d} |")
460
+ for k, l in zip(OP_KEYS, OP_LABELS_UI):
461
  if k in perturb_results:
462
  m = perturb_results[k]; d = f"{m['auc']-bl_auc:+.4f}"
463
  rows.append(f"| {l} | 推理期 | {m['auc']:.4f} | {m['attack_accuracy']:.4f} | {m['precision']:.4f} | {m['recall']:.4f} | {m['f1']:.4f} | {m['tpr_at_5fpr']:.4f} | {m['tpr_at_1fpr']:.4f} | {m['loss_gap']:.4f} | {bl_acc:.1f}% | {d} |")
 
540
  """)
541
 
542
  # 实验体系总览图 (如果在目录里则显示)
543
+ if os.path.exists(os.path.join(BASE_DIR, "figures", "algo4_overview_cn_final.png")):
544
+ gr.Image(os.path.join(BASE_DIR, "figures", "algo4_overview_cn_final.png"), label="实验体系总览", show_label=True)
545
 
546
  gr.HTML(f"""<div style="display:grid;grid-template-columns:repeat(4,1fr);gap:20px;margin:30px 0;">
547
  <div class="card-wrap" style="text-align:center;">
 
654
  gr.Markdown(f"### 1️⃣ 攻击成功率全景对比 (AUC)\n\n> 柱子越短 = AUC越低 = 防御越有效。基线AUC={bl_auc:.4f},标签平滑最低降至{gm('smooth_eps_0.2','auc'):.4f},输出扰动最低降至{gm('perturbation_0.03','auc'):.4f}。")
655
  gr.Plot(value=fig_auc_bar())
656
 
 
657
  gr.Markdown(f"""\
658
  ### 2️⃣ 多指标雷达图对比(全部11组实验)
659
 
660
  > **左图:标签平滑系列5个模型**
661
  > - 红色(Baseline)面积最大 = 攻击全面有效
662
+ > - 随着 ε 0.02 增至 0.2,雷达面积逐步缩小 = 防御逐步增强
663
  > - 特别注意 TPR@1%FPR 和 LossGap 两个轴,缩小最显著
664
  >
665
  > **右图:输出扰动系列7个配置**
666
  > - 红色(Baseline)同样是最大的
667
+ > - 随着 σ 0.005 增至 0.03,绿色系雷达逐步缩小
668
  > - OP在LossGap和TPR@5%维度上降幅尤其明显
669
  >
670
  > **结论:** 两种防御均在所有维度上全面压制攻击能力,不是只降低了某一个指标。
671
  """)
672
  gr.Plot(value=fig_radar())
 
673
 
674
  gr.Markdown("### 3️⃣ ROC曲线对比\n\n> 曲线越贴近对角线=攻击越接近随机猜测=防御越有效。左图标签平滑,右图输出扰动。")
675
  gr.Plot(value=fig_roc_curves())
 
795
 
796
  """)
797
 
798
+ demo.launch()