How Spam Filters Shaped the Field of Adversarial ML

This story was originally published on HackerNoon at: https://hackernoon.com/how-spam-filters-shaped-the-field-of-adversarial-ml. Evasion attacks and data poisoning let spammers bypass filters, turning the early-2000s inbox into a lab that shaped adversarial machine learning. Check more stories related to cybersecurity at: https://hackernoon.com/c/cybersecurity. You can also check exclusive content about #ai-security, #adversarial-machine-learning, #data-poisoning, #bayesian-spam-filtering, #the-history-of-spam-filters, #ml-evasion-techniques, #spam-detection-algorithms, #hackernoon-top-story, and more. This story was written by: @gthmk. Learn more about this writer by checking @gthmk's about page, and for more stories, please visit hackernoon.com. The 2000s spam arms race was an early stress test for adversarial ML. Spammers learned to manipulate inputs without seeing the model, close feedback loops with tracking pixels, and poison training data with as little as 1% corrupted samples. Every one of those attacks has a modern descendant in today's AI systems. The lesson the spam arms race exposed still holds: accuracy alone is not a sufficient measure of performance when an adversary can manipulate both model inputs and training data.