End-to-End (E2E) Reinforcement Learning (RL) for autonomous driving still struggles with safety and generalization under distribution shift, as perception-heavy encoders, sparse rewards, and ad hoc uncertainty handling often yield brittle closed-loop behavior. This work introduces a unified Deep RL (DRL) framework addressing key gaps: causal ego-centric state design, dense differentiable rewards, joint uncertainty estimation with entropy gating, and control-level policy transfer. An ego-centric relational graph encodes agent influence via uncertainty-weighted attention over kinematics, lane geometry, and semantics, producing a compact control state. A multi-objective differentiable reward stabilizes optimization by shaping safety, progress, and comfort with an uncertainty term. Aleatoric and epistemic uncertainty-captured through per-edge heteroscedastic variance and a critic ensemble-are aggregated into a calibrated confidence signal that modulates policy entropy for risk-aware exploration. A causal-semantic transfer objective aligns actions, attention, and uncertainty statistics across domains, combined with meta-learned initialization for few-shot adaptation. In closed-loop urban driving across varied towns, traffic, and weather, the framework improves success rate, reduces infractions per kilometer, and achieves higher time-to-conflict with lower lateral deviation and comfort cost compared to strong baselines.

Personal RFID tags store valuable information private to their users that can easily be subject to eavesdropping, unauthorized reading, owner tracking, and cloning. RFID tags are also susceptible to relay attacks and likely to get lost and stolen. In this paper, we introduce the problem of user authentication to RFID tags. This allows users to control when and where their RFID tags can be accessed. We present a novel approach for user authentication to multiple RFID tags called "Vibrate-to-Unlock" (VtU). This technique uses a mobile phone as an authentication token, forming an unidirectional tactile communication channel between users and their RFID tags. Authenticating to an RFID tag involves touching a vibrating phone to the tag or an object carrying the tag, such as a wallet. We discuss the design and implementation of this new method on Intel's WISP tags. We also report on a preliminary usability evaluation of our VtU prototype.

Clock skew, an inherent property of clock crystals of physical devices, is defined as the rate of deviation of a device clock from the true time. The frequency of a device's clock actually depends on its environment, such as the temperature, humidity, vibration, electromagnetic interference, as well as the type of crystal. The main contributions of this paper are twofold. First, we experimentally validate that MICAz and TelosB sensor motes have different and unique clock skews. Furthermore, the clock skew of a node can easily be monitored, even via a multi-hop Wireless Sensor Network (WSN). We argue that this feature can be used for identification of the nodes, detection of wormhole and Sybil attacks. Second, we show that the clock skew of a sensor node varies with the variation of temperature. We explain how this property can be used to detect malicious and malfunctioning nodes and to geolocalise them. We also discuss the pros and cons of utilisation of the above two properties for different services in WSNs.

“Secure Device Pairing” is the process of bootstrapping secure communication between two devices over a short- or medium-range wireless channel (such as Bluetooth, WiFi). The devices in such a scenario can neither be assumed to have a prior context with each other nor do they share a common trusted authority. Fortunately, the devices can generally be connected using auxiliary physical channel(s) (such as audio, visual, tactile) that can be authenticated by the device user(s), thus forming the basis for pairing. However, lack of good quality output interfaces (e.g, a speaker, display) and/or receivers (e.g., microphone, camera) on certain devices makes pairing a very challenging problem in practice. We consider the problem of “rushing user” behavior in device pairing. A rushing user is defined as a user who in a rush to connect her devices, would skip through the pairing process, if possible. Most prior pairing methods, in which the user decides the final outcome of pairing, are vulnerable to rushing user behavior – the user can simply “accept” the pairing, without having to correctly take part in the decision process. In this paper, we concentrate on most common pairing scenarios (such as pairing of a WiFi laptop and an access point), whereby one device (access point) is constrained in terms output interfaces, while the other (laptop) has a decent quality output interface but no receiver. We present the design and usability analysis of two novel pairing methods, which are resistant to a rushing user and require only minimal device interfaces on the constrained device. One of the most appealing applications of our proposal is in defending against common threat of “Evil Twin” attacks in public places (e.g, cyber-cafes, airport lounges).

Wireless sensor networks have several useful applications in commercial and defense settings, as well as user-centric personal area networks. To establish secure (point-to-point and/or broadcast) communication channels among the nodes of a wireless sensor network is a fundamental security task. To this end, a plethora of so-called key pre-distribution schemes have been proposed in the past, e.g., [25][9][19][8][5]. All these schemes, however, rely on shared secret(s), which are assumed to be pre-loaded onto the sensor nodes, e.g., during the manufacturing process. In this paper, we consider the problem of user-assisted secure initialization of sensor network necessary to bootstrap key pre-distribution. This is a challenging problem due to the level of user burden involved in initializing multiple (often large number of) sensor nodes and lack of input and output user-interfaces on sensor motes. We propose a novel method for secure sensor node initialization based on a visual out-of-band channel that utilizes minimal output interface in the form of LED(s) already available on most off-the-shelf sensor motes. The proposed method requires only a little extra cost, is efficient and reasonably scalable. Moreover, based on a usability study that we conducted, the method turns out to be quite user-friendly and easy to administer by everyday computer users.

User-to-tag authentication can prevent a variety of potential attacks on personal RFID tags. In this poster, a new RFID authentication scheme is presented that allows a user to control when a tag responds to queries by leveraging a mobile phone. The design and implementation of this approach is presented along with a study of its usability.

The operation of achieving authenticated key agreement between two human-operated devices over a short-range wireless communication channel (such as Bluetooth or WiFi) is referred to as "Pairing". The devices in such a scenario are ad hoc in nature, i.e., they can neither be assumed to have a prior context (such as pre-shared secrets) with each other nor do they share a common trusted on- or off-line authority. However, the devices can generally be connected using auxiliary physical channel(s) (such as audio, visual, etc.) that can be authenticated by the device user(s) and thus form a basis for pairing. One of the main challenges of secure device pairing is the lack of good quality output interfaces as well as corresponding receivers on devices. In [13], we presented a pairing scheme which is universally applicable to any pair of devices (such as a WiFi AP and a laptop, a Bluetooth keyboard and a desktop, etc.). The scheme is based upon the device user(s) comparing short and simple synchronized audiovisual patterns, such as "beeping" and "blinking". In this paper, we automate the (manual) scheme of [13] by making use of an auxiliary, commonly available device such as a personal camera phone. Based on a preliminary user study we conducted, we show that the automated scheme is generally faster and more user-friendly relative to the manual scheme. More importantly, the proposed scheme turns out to be quite accurate in the detection of any possible attacks.

To establish secure (point-to-point and/or broadcast) communication channels among the nodes of a wireless sensor network is a fundamental task. To this end, a plethora of (socalled) key pre-distribution schemes have been proposed in the past. All these schemes, however, rely on shared secret(s), which are assumed to be somehow pre-loaded onto the sensor nodes. In this paper, we propose a novel method for secure initialization of sensor nodes based on a visual out-of-band channel. Using the proposed method, the administrator of a sensor network can distribute keys onto the sensor nodes, necessary to bootstrap key pre-distribution. Our secure initialization method requires only a little extra cost, is efficient and scalable with respect to the number of sensor nodes. Moreover, based on a usability study that we conducted, the method turns out to be quite user-friendly and easy to use by naive human users.

"Secure Device Pairing" is the process of bootstrapping secure communication between two human-operated devices over a short- or medium-range wireless channel (such as Bluetooth, WiFi). The devices in such a scenario can neither be assumed to have a prior context with each other nor do they share a common trusted authority. However, the devices can generally be connected using auxiliary physical channel(s) (such as audio, visual) that can be authenticated by the device user(s), and thus form the basis for pairing.Recently proposed pairing protocols are based upon bidirectional physical channels. However, various pairing scenarios are asymmetric in nature, i.e., only a unidirectional physical channel exists between two devices (such as between a cell phone and an access point). In this paper, we concentrate on pairing devices using a unidirectional physical channel and analyze recently proposed protocol on this topic [14]. Moreover, as an improvement to [14], we present an efficient implementation of a unidirectional physical channel based on multiple blinking LEDs as transmitter and a video camera as a receiver.

"Pairing" is referred to as the operation of achieving authenticated key agreement between two human-operated devices over a short-range wireless communication channel (such as Bluetooth, WiFi). The devices are ad hoc in nature, i.e., they can neither be assumed to have a prior context with each other nor do they share a common trusted authority. However, the devices can generally be connected using auxiliary physical channel(s) (such as audio. Visual, etc.) that can be authenticated by the device user(s), and thus form the basis for pairing.Recently proposed pairing protocols are based upon bidirectional physical channels. However, in various pairing scenarios, only a unidirectional physical channel exists between two devices (such as between an access point and a cell phone). In this paper, we consider pairing devices using a unidirectional physical channel and focus on a recently proposed protocol on this topic by Saxena et al. [13]. As an improvement to [13], we present an efficient implementation of a physical channel based on multiple LEDs as transmitters and a video camera as a receiver.

An attempt has been made in this paper to estimate the reliability of an s-out-of-k system with non-identical component strengths when component strengths follow an exponential distribution. A further assumption is made that all the components are subjected to a common random stress which also follow an exponential distribution. Bayes and maximum likelihood estimators of such system reliability are considered. A Bayes estimate is obtained by using Lindley's approximation. Comparisons are made on the basis of efficiency and Pitman nearness probability through a Monte-Carlo study.

In this paper we present three models for the behavior of software failures. By applying these models an attempt has been made to predict reliability growth by predicting failure rates and mean time to next failure of software with Weibull inter failure times at different stages. The changes in the performance of the software as a result of the error removal are described as a Bayes empirical-Bayes prediction in Model I. Model II considers a fully Bayesian analysis with non informative priority of Weibull parameters. An approximation due to Lindley is used in this model as the expressions do not appear in closed forms. The M.L. approach is used in Model III. Finally we apply these three models to actual failure data and compare their predictive performances. The comparison of the proposed models is also made in terms of the ratio of likelihoods of observed values based on their predictive distributions.Among these three models, Model I seems to be quite reasonable as it shows higher reliability growth in ail stages. It is noted that this model may be useful to measure the current reliability at any particular stage of the testing process and viewed as a measure of software quality.

This paper draws inferences regarding the reliability in a multicomponent stress-strength system when both stress and strength are independently identically Burr random variables. Maximum likelihood (ML) and Bayes estimators of the system reliability are considered. A numerical example is presented on the basis of a Monte Carlo simulation.

A multi-component stress-strength model of an s-out-of-k system is considered. Johnson (Handbook of Statistics, vol. 7. Elsevier Science Publishers, 1988, pp. 27–54) introduced the generalization of such a system by considering a non-identical component's strength distribution and also found the maximum likelihood estimate (mle) by considering an exponential distribution of stress and strength. Bayes estimate of such a system's reliability function is obtained by using the Lindley's method of approximation (Trabajos de Estadisticay Investiracion Operative, 31 (1980) 232–245). The component strengths follow independent but not all identical Weibull distributions. It is further assumed that all the components are subjected to a common random stress which is also distributed as a Weibull random variable. A squared error loss function is used. A numerical example is presented in which comparison is made with the mle obtained by a Monte Carlo study of efficiency and Pitman nearness probability.

This paper draws inferences about the reliability in a multi-component stress-strength system when both stress and strength are independently identically distributed (idd) Burr random variables. We consider both maximum likelihood and Bayes estimators of the system reliability. The two estimators are compared numerically by obtaining empirical efficiencies with respect to the maximum likelihood estimator (MLE) by generating 1000 random samples by a Monte Carlo simulation. It is found that the Bayes estimators are better than the corresponding MLEs for small samples (ni ≤ 7; i = 1, 2). Moreover, the robustness of the Bayes estimators to the change of the prior parameters is also considered.

A procedure for choosing between models from two families of distributions for failure data is given. This method is based on F-statistics and can be used for a non-censored set of life testing observations. Monte Carlo samples from exponential , Wei bull , Pareto and finite range distributions are used. The performance of the proposed procedure is compared to the selection procedure S, derived by Quesenberry and Kent (1982)and to the MLR test considered by Bain and Engelhardt (l980).