Biblio

In recent years, behavioral biometrics have become a popular approach to support continuous authentication systems. Most generally, a continuous authentication system can make two types of errors: false rejects and false accepts. Based on this, the most commonly reported metrics to evaluate systems are the False Reject Rate (FRR) and False Accept Rate (FAR). However, most papers only report the mean of these measures with little attention paid to their distribution. This is problematic as systematic errors allow attackers to perpetually escape detection while random errors are less severe. Using 16 biometric datasets we show that these systematic errors are very common in the wild. We show that some biometrics (such as eye movements) are particularly prone to systematic errors, while others (such as touchscreen inputs) show more even error distributions. Our results also show that the inclusion of some distinctive features lowers average error rates but significantly increases the prevalence of systematic errors. As such, blind optimization of the mean EER (through feature engineering or selection) can sometimes lead to lower security. Following this result we propose the Gini Coefficient (GC) as an additional metric to accurately capture different error distributions. We demonstrate the usefulness of this measure both to compare different systems and to guide researchers during feature selection. In addition to the selection of features and classifiers, some non- functional machine learning methodologies also affect error rates. The most notable examples of this are the selection of training data and the attacker model used to develop the negative class. 13 out of the 25 papers we analyzed either include imposter data in the negative class or randomly sample training data from the entire dataset, with a further 6 not giving any information on the methodology used. Using real-world data we show that both of these decisions lead to significant underestimation of error rates by 63% and 81%, respectively. This is an alarming result, as it suggests that researchers are either unaware of the magnitude of these effects or might even be purposefully attempting to over-optimize their EER without actually improving the system.

In recent years, behavioral biometrics have become a popular approach to support continuous authentication systems. Most generally, a continuous authentication system can make two types of errors: false rejects and false accepts. Based on this, the most commonly reported metrics to evaluate systems are the False Reject Rate (FRR) and False Accept Rate (FAR). However, most papers only report the mean of these measures with little attention paid to their distribution. This is problematic as systematic errors allow attackers to perpetually escape detection while random errors are less severe. Using 16 biometric datasets we show that these systematic errors are very common in the wild. We show that some biometrics (such as eye movements) are particularly prone to systematic errors, while others (such as touchscreen inputs) show more even error distributions. Our results also show that the inclusion of some distinctive features lowers average error rates but significantly increases the prevalence of systematic errors. As such, blind optimization of the mean EER (through feature engineering or selection) can sometimes lead to lower security. Following this result we propose the Gini Coefficient (GC) as an additional metric to accurately capture different error distributions. We demonstrate the usefulness of this measure both to compare different systems and to guide researchers during feature selection. In addition to the selection of features and classifiers, some non- functional machine learning methodologies also affect error rates. The most notable examples of this are the selection of training data and the attacker model used to develop the negative class. 13 out of the 25 papers we analyzed either include imposter data in the negative class or randomly sample training data from the entire dataset, with a further 6 not giving any information on the methodology used. Using real-world data we show that both of these decisions lead to significant underestimation of error rates by 63% and 81%, respectively. This is an alarming result, as it suggests that researchers are either unaware of the magnitude of these effects or might even be purposefully attempting to over-optimize their EER without actually improving the system.

Growing numbers of ubiquitous electronic devices and services motivate the need for effortless user authentication and identification. While biometrics are a natural means of achieving these goals, their use poses privacy risks, due mainly to the difficulty of preventing theft and abuse of biometric data. One way to minimize information leakage is to derive biometric keys from users' raw biometric measurements. Such keys can be used in subsequent security protocols and ensure that no sensitive biometric data needs to be transmitted or permanently stored. This paper is the first attempt to explore the use of human body impedance as a biometric trait for deriving secret keys. Building upon Randomized Biometric Templates as a key generation scheme, we devise a mechanism that supports consistent regeneration of unique keys from users' impedance measurements. The underlying set of biometric features are found using a feature learning technique based on Siamese networks. Compared to prior feature extraction methods, the proposed technique offers significantly improved recognition rates in the context of key generation. Besides computing experimental error rates, we tailor a known key guessing approach specifically to the used key generation scheme and assess security provided by the resulting keys. We give a very conservative estimate of the number of guesses an adversary must make to find a correct key. Results show that the proposed key generation approach produces keys comparable to those obtained by similar methods based on other biometrics.

Security of embedded devices is a timely and important issue, due to the proliferation of these devices into numerous and diverse settings, as well as their growing popularity as attack targets, especially, via remote malware infestations. One important defense mechanism is remote attestation, whereby a trusted, and possibly remote, party (verifier) checks the internal state of an untrusted, and potentially compromised, device (prover). Despite much prior work, remote attestation remains a vibrant research topic. However, most attestation schemes naturally focus on the scenario where the verifier is trusted and the prover is not. The opposite setting–-where the prover is benign, and the verifier is malicious–-has been side-stepped. To this end, this paper considers the issue of prover security, including: verifier impersonation, denial-of-service (DoS) and replay attacks, all of which result in unauthorized invocation of attestation functionality on the prover. We argue that protection of the prover from these attacks must be treated as an important component of any remote attestation method. We formulate a new roaming adversary model for this scenario and present the trade-offs involved in countering this threat. We also identify new features and methods needed to protect the prover with minimal additional requirements.