SecAlign
A fine-tuning-based defense method against LLM prompt injection that adjusts model behavior through training to resist injection attacks, but may impair the model's general performance in settings such as few-shot scenarios.
Timeline (last 90 days)
Under the black-box RL-Hammer attack, HoS's attack success rate (ASR) is only 10%, compared with 97% for the original GPT-OSS-20B model and 85% for SecAlign.
HoS maintains the base model's performance on standard multiple-choice benchmarks and prompt settings, whereas SecAlign and ASIDE degrade in few-shot scenarios.
On agent benchmarks of static attacks, HoS and SecAlign perform similarly and substantially outperform the safety-utility trade-off of base models.
All Facts (3)
Under the black-box RL-Hammer attack, HoS's attack success rate (ASR) is only 10%, compared with 97% for the original GPT-OSS-20B model and 85% for SecAlign.
50%UnverifiedHoS maintains the base model's performance on standard multiple-choice benchmarks and prompt settings, whereas SecAlign and ASIDE degrade in few-shot scenarios.
50%UnverifiedOn agent benchmarks of static attacks, HoS and SecAlign perform similarly and substantially outperform the safety-utility trade-off of base models.
50%