Human Interaction Proofs

昨天上午與學生 meeting, 由於小和的實驗上沒有具體進度, 所以我們就針對微軟所發表的一篇論文討論。這篇論文的首位作者是 Kumar Chellapilla, 任職於 Microsoft Research, 研究主題之一就是 Human Interaction Proof, 至今一共發表了四篇論文, 昨天我們討論的論文是:
Building Segmentation Based Human-friendly Human Interaction Proofs (HIPs)

所謂的 Human Interaction Proof (HIP) 就是在 Internet 上, 用來證明 (prove) 某個互動 (interaction) 是由人 (human) 所發出的, 而不是由程式自動發出的。

大家要在回應文章時, 如果沒有登入自己的帳號, 系統會要求你輸入某個圖片中的文數字, 當你輸入正確時, 就代表系統認為這個回應是人發出的, 而不是網路電腦程式自動發出的。這樣做的目的就是希望阻止一些廣告或惡意癱瘓系統的程式利用大量回應的方式, 達到廣告或癱瘓系統的目的。

這篇論文的摘要寫得非常好, 完全符合科學論文摘要的要求, 沒有多寫一句話, 也沒有少寫一句話, 重點都寫出來了, 把 HIP 的重點與考量都交代的一清二楚, 同學寫論文時, 可以參考。

Abstract
Human interaction proofs (HIPs) have become common place on the internet due to their effectiveness in deterring automated abuse of online services intended for humans. However, there is a co-evolutionary arms race in progress and these proofs are becoming more difficult for genuine users while attackers are getting better at breaking existing HIPs. We studied various popular HIPs on the internet to understand their strength and human friendliness. To determine HIP strength, we adopted a direct approach of building computer attacks using image processing and machine learning techniques. To understand human-friendliness, a sequence of users studies were conducted to investigate HIP character recognition by humans under a variety of visual distortions and clutter commonly employed in reading-based HIPs. We found that many of the online HIPs are pure recognition tasks that can be easily broken using machine learning. The stronger HIPs tend to pose a combination of segmentation and recognition challenges. Further, the HIP user studies show that given correct segmentation, computers are much better at HIP character recognition than humans.

摘要的第一句話說明HIPs 變得越來越常見的原因。
Human interaction proofs (HIPs) have become common place on the internet due to their effectiveness in deterring automated abuse of online services intended for humans.

第二句, 用 however 來表達話風一轉的語氣, 馬上指出 HIPs 所面臨的困難, 這句話也引出作者想要解決的問題。
However, there is a co-evolutionary arms race in progress and these proofs are becoming more difficult for genuine users while attackers are getting better at breaking existing HIPs.

第三句話說明了作者藉由研究許多 HIPs 系統, 了解到一個好的 HIPs 需要同時去考量 security strength 與 human friendliness 兩個因素。
We studied various popular HIPs on the internet to understand their strength and human friendliness.

接下來都是用不定詞 To 所引領的兩句話, 就是分別說明作者是如何考量這兩個因素。

To determine HIP strength, we adopted a direct approach of building computer attacks using image processing and machine learning techniques.

To understand human-friendliness, a sequence of users studies were conducted to investigate HIP character recognition by humans under a variety of visual distortions and clutter commonly employed in reading-based HIPs.

然後, 作者開始說明研究成果為何。
We found that many of the online HIPs are pure recognition tasks that can be easily broken using machine learning.
The stronger HIPs tend to pose a combination of segmentation and recognition challenges.
Further, the HIP user studies show that given correct segmentation, computers are much better at HIP character recognition than humans.