Fnu Suya

Fine-tuned TTS models expose severe privacy risks, with membership inference attacks achieving speaker-level AUCs nearing perfect accuracy.

A single parasitic Trojan can turn benign data poisoning into a 100% effective backdoor attack by leveraging real-time interactions within ML training dynamics.

LLM providers can inflate your bill by over 14x without detection, thanks to unauditable token counts and a "trust paradox" at the heart of current auditing frameworks.