Contoh Permusuhan dan Kuis Kekokohan

Jawaban dan penjelasan muncul saat Anda mengikuti kuis.

1. Under a specified threat model, what is an adversarial example?
   • An input crafted to cause a model error using changes allowed by that threat model
   • Any ordinary prediction error on unmodified input
   • An input that must contain tiny, imperceptible pixel changes
   • A model update that improves accuracy on every distribution
2. Apa yang membedakan serangan 'kotak putih' dengan serangan 'kotak hitam'?
   • Serangan kotak putih hanya bekerja pada gambar berwarna putih
   • Di kotak putih, penyerang memiliki akses ke internal/gradien model; di kotak hitam, hanya keluaran
   • Serangan kotak hitam selalu lebih cepat
   • Serangan kotak putih tidak memerlukan komputasi
3. FGSM (Metode Tanda Gradien Cepat) membuat contoh permusuhan dengan...
   • Mengacak piksel secara acak
   • Melatih ulang model dari awal
   • Mengambil langkah searah dengan tanda gradien kerugian terhadap masukan
   • Memotong gambar
4. What is adversarial training commonly used for?
   • Guaranteeing robustness to every possible attack
   • Proving a classifier is fair
   • Eliminating all clean-data errors
   • Improving robustness against the adversarial threat model represented during training
5. Mengapa 'keteralihan' contoh-contoh yang bermusuhan menjadi perhatian?
   • Serangan yang dilakukan pada satu model sering kali menipu model lain, sehingga memungkinkan terjadinya serangan kotak hitam
   • Itu membuat model berlatih lebih lambat
   • Ini meningkatkan biaya penyimpanan
   • Ini hanya mempengaruhi model audio
6. How are prompt-injection attacks related to image adversarial examples?
   • They are identical bounded pixel perturbations
   • Both exploit model vulnerabilities, but prompt injection has a distinct instruction and system threat model
   • Prompt injection requires changing model weights
   • Image adversarial examples only affect text prompts
7. Why must an adversarial-robustness report specify the perturbation norm and budget?
   • They determine the model’s number of parameters
   • They guarantee transfer to every attack
   • They define which input changes the reported robustness claim covers
   • They replace the need to state an attacker’s access
8. Which pair of measures helps distinguish clean performance from performance under a specified attack?
   • Training-set size and GPU memory
   • Tokenization rate and context length
   • Model file size and parameter dtype
   • Clean accuracy and robust accuracy under the stated attack setting