Kurt Alfred Kluever

Kurt Alfred Kluever

Kurt Alfred Kluever graduated from the Rochester Institute of Technology in 2008, where he received both a BS and MS in Computer Science. He is currently a Software Engineer at Google New York. His research interests include CAPTCHAs, web security, pattern recognition, and machine learning.
Authored Publications
Google Publications
Other Publications
Sort By
  • Title
  • Title, descending
  • Year
  • Year, descending
    Balancing Usability and Security in a Video CAPTCHA
    Richard Zanibbi
    Proceedings of the 5th Symposium on Usable Privacy and Security (SOUPS '09), ACM Press(2009)
    Preview abstract We present a technique for using a content-based video labeling task as a CAPTCHA. Our video CAPTCHAs are generated from YouTube videos, which contain labels (tags) supplied by the person that uploaded the video. They are graded using a video's tags, as well as tags from related videos. In a user study involving 184 participants, we were able to increase the average human success rate on our video CAPTCHA from roughly 70% to 90%, while keeping the average success rate of a tag frequency-based attack fixed at around 13%. Through a different parameterization of the challenge generation and grading algorithms, we were able to reduce the success rate of the same attack to 2%, while still increasing the human success rate from 70% to 75%. The usability and security of our video CAPTCHA appears to be comparable to existing CAPTCHAs, and a majority of participants (60%) indicated that they found the video CAPTCHAs more enjoyable than traditional CAPTCHAs in which distorted text must be transcribed. View details
    Video CAPTCHAs: Usability vs. Security
    Richard Zanibbi
    Proceedings of the IEEE Western New York Image Processing Workshop (WNYIP '08), IEEE Press(2008)
    Preview abstract A CAPTCHA is a variation of the Turing test, in which a challenge is used to distinguish humans from computers (”bots”) on the internet. They are commonly used to prevent the abuse of online services. CAPTCHAs discriminate using hard artificial intelligence problems: the most common type requires a user to transcribe distorted characters displayed within a noisy image. Unfortunately, many users find them frustrating and break rates as high as 60% have been reported (for Microsoft’s Hotmail). We present a new CAPTCHA in which users provide three words (”tags”) that describe a video. A challenge is passed if a user’s tag belongs to a set of automatically generated ground-truth tags. In an experiment, we were able to increase human pass rates for our video CAPTCHAs from 69.7% to 90.2% (184 participants over 20 videos). Under the same conditions, the pass rate for an attack submitting the three most frequent tags (estimated over 86,368 videos) remained nearly constant (5% over the 20 videos, roughly 12.9% over a separate sample of 5146 videos). Challenge videos were taken from YouTube.com. For each video, 90 tags were added from related videos to the ground-truth set; security was maintained by pruning all tags with a frequency ≥ 0.6%. Tag stemming and approximate matching were also used to increase human pass rates. Only 20.1% of participants preferred text-based CAPTCHAs, while 58.2% preferred our video-based alternative. Finally, we demonstrate how our technique for extending the ground truth tags allows for different usability/security trade-offs, and discuss how it can be applied to other types of CAPTCHAs. View details
    Breaking the PayPal.com HIP: A Comparison of Classifiers
    Rochester Institute of Technology(2008)
    Preview abstract Human Interactive Proofs (HIPs) are a method used to differentiate between humans and machines on the internet. Providers of online services such as PayPal.com use HIPs to prevent automated signups and abuse of their services. In this experiment, a three step algorithm has been developed to break the PayPal.com HIP. The image is preprocessed to remove noise using thresholding and a simple cleaning technique, and then segmented using vertical projections and candidate split positions. Four classification methods have been implemented: pixel counting, vertical projections, horizontal projections and template correlations. The system was trained on a sample of twenty PayPal.com HIPs to create thirty-six training templates (one for each character: 0-9 and A-Z). A sample of 100 PayPal.com HIPs were used for testing. The following HIP success rates have been achieved using the different classifiers: 8% pixel counting, vertical projections 97%, horizontal projections 100%, template correlations 100%. Three of the classifers out perform the 88% HIP success rate of [6]. View details
    Evaluating the Usability and Security of a Video CAPTCHA
    Rochester Institute of Technology(2008)
    Preview abstract A CAPTCHA is a variation of the Turing test, in which a challenge is used to distinguish humans from computers ("bots") on the internet. They are commonly used to prevent the abuse of online services. CAPTCHAs discriminate using hard artificial intelligence problems: the most common type requires a user to transcribe distorted characters displayed within a noisy image. Unfortunately, many users find them frustrating and break rates as high as 60% have been reported (for Microsoft’s Hotmail). We present a new CAPTCHA in which users provide three words ("tags") that describe a video. A challenge is passed if a user’s tag belongs to a set of automatically generated ground-truth tags. In an experiment, we were able to increase human pass rates for our video CAPTCHAs from 69.7% to 90.2% (184 participants over 20 videos). Under the same conditions, the pass rate for an attack submitting the three most frequent tags (estimated over 86,368 videos) remained nearly constant (5% over the 20 videos, roughly 12.9% over a separate sample of 5146 videos). Challenge videos were taken from YouTube.com. For each video, 90 tags were added from related videos to the ground-truth set; security was maintained by pruning all tags with a frequency ≥ 0.6%. Tag stemming and approximate matching were also used to increase human pass rates. Only 20.1% of participants preferred text-based CAPTCHAs, while 58.2% preferred our video-based alternative. Finally, we demonstrate how our technique for extending the ground truth tags allows for different usability/security trade-offs, and discuss how it can be applied to other types of CAPTCHAs. View details
    Improving Measurement Accuracy in Sensor Networks by an Object Model Generation and Application
    Leonid Reznik
    Proceedings of the 6th Annual IEEE Conference on Sensors (SENSORS '07), IEEE Press, Atlanta, GA(2007), pp. 371-374
    Preview abstract The paper describes a novel method of calculating measurement results in sensor networks, which includes modifying the conventional measurement estimates based on the object under measurement model mined from the data collected by the sensor network itself previously and other information made available by domain experts. It is shown that the model application might produce a significant gain in measurement accuracy if the model is correct. The gain value is estimated and its dependence on various factors is studied by computer simulation and experimentation with real sensor networks built from Crossbow Telos ver. B motes. The conditions of achieving the gain versus suffering the loss are derived and the recommendations of how to shape the object model in order to achieve and maximize the gain value are provided. View details