CAPTCHA (or Human Interaction Proof) is now almost a standard security technique for defending against undesirable or malicious bot programs on the Internet. However, the robustness of CAPTCHAs has so far been studied mainly just in communities such as computer vision, and document analysis and recognition. This paper motivates a security engineering perspective of the robustness of CAPTCHAs. Specifically, we show that a number of CAPTCHAs that appeared to be secure, including schemes widely deployed by Microsoft, Yahoo and Google and some other less well-known ones, could be broken with a high success rate with simple but novel attacks. In contrast to earlier work that relied on sophisticated computer vision algorithms, our attacks exploited critical design errors that we discovered in each scheme. The main lesson is that security engineering expertise and experience, in particular adversarial thinking skills, can make a unique and significant contribution to the improvement of the robustness of CAPTCHAs.

In the 198Os, considerable advances were made in both software and hardware technology, and CPUs that can issue no more than one operation per clock cycle are rapidly approaching this barrier. Further improvements to uniprocessor performance can be obtained by enhancing the architecture of the CPU to allow multiple operations to be issued in a single clock cycle. The focus of this panel is to discuss three architectural approaches to issuing multiple operations per cycle: 6) vector instructions, (ii) very long instruction words (VLIW), and (iii) superscalar execution. We have asked the panelists to address several issues from the vantage point of their preferred architectural approach. These include: