Examples
We present examples accompaning our USENIX Security'23 paper “No more Reviewer #2: Subverting Automatic Paper-Reviewer Assignment using Adversarial Learning”. We show three representative adversarial papers computed for different scenarios and configurations.No more Reviewer #2: Subverting Automatic Paper-Reviewer Assignment using Adversarial Learning
ArXiv ID: 2303.14443Transformations (Budget n/a)
Hidden Box
Selected Reviewer
Reviewer 4, Reviewer 7
Rejected Reviewer
Reviewer 5
Ranking
Original Ranking
| Rank | Reviewer | Score | ||
|---|---|---|---|---|
| 1 | Reviewer 1 | 1.00 | ||
| 2 | Reviewer 2 | 0.99 | ||
| 3 | Reviewer 3 | 0.95 | ||
| 4 | Reviewer 4 | 0.87 | ||
| 5 | Reviewer 5 | 0.80 | ||
| 6 | Reviewer 6 | 0.73 | ||
| 7 | Reviewer 7 | 0.61 | ||
| 8 | Reviewer 8 | 0.60 | ||
| 9 | Reviewer 9 | 0.58 | ||
| 10 | Reviewer 10 | 0.58 |
Adversarial Ranking
| Rank | Reviewer | Score | ||
|---|---|---|---|---|
| 1 | Reviewer 2 | 1.00 | ||
| 2 | Reviewer 1 | 1.00 | ||
| 3 | Reviewer 3 | 0.96 | ||
| 4 | Reviewer 4 | 0.91 | ||
| 5 | Reviewer 7 | 0.89 | ||
| 6 | Reviewer 8 | 0.81 | ||
| 7 | Reviewer 5 | 0.77 | ||
| 8 | Reviewer 9 | 0.75 | ||
| 9 | Reviewer 6 | 0.74 | ||
| 10 | Reviewer 10 | 0.68 |
Backdooring and Poisoning Neural Networks with Image-Scaling Attacks
ArXiv ID: 2003.08633Transformations (Budget n/a)
Hidden Box
Selected Reviewer
Reviewer 1, Reviewer 2, Reviewer 4, Reviewer 6, Reviewer 10
Rejected Reviewer
None
Ranking
Original Ranking
| Rank | Reviewer | Score | ||
|---|---|---|---|---|
| 1 | Reviewer 1 | 1.00 | ||
| 2 | Reviewer 2 | 0.94 | ||
| 3 | Reviewer 3 | 0.93 | ||
| 4 | Reviewer 4 | 0.87 | ||
| 5 | Reviewer 5 | 0.83 | ||
| 6 | Reviewer 6 | 0.68 | ||
| 7 | Reviewer 7 | 0.51 | ||
| 8 | Reviewer 8 | 0.50 | ||
| 9 | Reviewer 9 | 0.45 | ||
| 10 | Reviewer 10 | 0.44 |
Adversarial Ranking
| Rank | Reviewer | Score | ||
|---|---|---|---|---|
| 1 | Reviewer 10 | 1.00 | ||
| 2 | Reviewer 1 | 0.99 | ||
| 3 | Reviewer 2 | 0.98 | ||
| 4 | Reviewer 4 | 0.97 | ||
| 5 | Reviewer 6 | 0.97 | ||
| 6 | Reviewer 3 | 0.96 | ||
| 7 | Reviewer 5 | 0.80 | ||
| 8 | Reviewer 7 | 0.79 | ||
| 9 | Reviewer 11 | 0.75 | ||
| 10 | Reviewer 12 | 0.69 |
Hardening Code Obfuscation Against Automated Attacks
ArXiv ID: 2106.08913Transformations (Budget 1.00)
Synonym, Language model, Reference addition
Selected Reviewer
Reviewer 7
Rejected Reviewer
Reviewer 5
Ranking
Original Ranking
| Rank | Reviewer | Score | ||
|---|---|---|---|---|
| 1 | Reviewer 1 | 1.00 | ||
| 2 | Reviewer 2 | 0.93 | ||
| 3 | Reviewer 3 | 0.83 | ||
| 4 | Reviewer 4 | 0.77 | ||
| 5 | Reviewer 5 | 0.76 | ||
| 6 | Reviewer 6 | 0.74 | ||
| 7 | Reviewer 7 | 0.73 | ||
| 8 | Reviewer 8 | 0.72 | ||
| 9 | Reviewer 9 | 0.70 | ||
| 10 | Reviewer 10 | 0.69 |
Adversarial Ranking
| Rank | Reviewer | Score | ||
|---|---|---|---|---|
| 1 | Reviewer 1 | 1.00 | ||
| 2 | Reviewer 2 | 0.92 | ||
| 3 | Reviewer 3 | 0.82 | ||
| 4 | Reviewer 4 | 0.77 | ||
| 5 | Reviewer 7 | 0.77 | ||
| 6 | Reviewer 8 | 0.74 | ||
| 7 | Reviewer 5 | 0.74 | ||
| 8 | Reviewer 6 | 0.72 | ||
| 9 | Reviewer 9 | 0.71 | ||
| 10 | Reviewer 10 | 0.69 |