Adversarial Attacks and Defenses in AI Systems: Challenges, Strategies, and Future Directions
Lawrence Samuel Igenewari, Onyemaechi Emmanuel Okoh · International Journal of Research and Innovation in Applied Science · 2025
AI systems are vulnerable to adversarial manipulations (Szegedy et al., 2014). These attacks exploit model weaknesses through subtle input perturbations (Carlini & Wagner, 2017), risking safety in applications like facial recognition and autonomous driving (Eykholt et al., 2018). Defense mechanisms, including adversarial training (Madry et al., 2018) and input preprocessing (Guo et al., 2018), often face trade-offs between robustness and efficiency.