Farah Saab

Energy efficient JPEG using stochastic processing

One of the serious challenges facing smartphones and other portable devices is battery life. We aim at improving the energy efficiency of multimedia applications for improved battery life while maintaining an acceptable level of user experience. In this paper, we design a stochastic processing technique for JPEG to limit its energy consumption particularly on mobile devices. This is a first step towards improving the energy efficiency for MPEG, since video streaming and conferencing applications are some of the most energy-consuming applications. We demonstrate the feasibility of using stochastic processing for multimedia encoders and analyze the resulting energy savings.

Solving sybil attacks using evolutionary game theory

Recommender systems have become quite popular recently. However, such systems are vulnerable to several types of attacks that target user ratings. One such attack is the Sybil attack where an entity masquerades as several identities with the intention of diverting user ratings. In this work, we propose evolutionary game theory as a possible solution to the Sybil attack in recommender systems. After modeling the attack, we use replicator dynamics to solve for evolutionary stable strategies. Our results show that under certain conditions that are easily achievable by a system administrator, the probability of an attack strategy drops to zero implying degraded fitness for Sybil nodes that eventually die out.

A crowdsourcing game-theoretic intrusion detection and rating system

One of the main concerns for smartphone users is the quality of apps they download. Before installing any app from the market, users first check its rating and reviews. However, these ratings are not computed by experts and most times are not associated with malicious behavior. In this work, we present an IDS/rating system based on a game theoretic model with crowdsourcing. Our results show that, with minor control over the error in categorizing users and the fraction of experts in the crowd, our system provides proper ratings while flagging all malicious apps.

Energy-Efficient Security for Voice over IP

The fast spread of handheld smart devices contributed to the development of VoIP softphones running over such devices. Most security mechanisms were mainly designed for desktop PCs and hence did not take into consideration the power constraints of handheld devices. This fact highly motivated the development of new security mechanisms that try to minimize the energy consumption without compromising the security of the exchanged data. In this paper, we propose an energy-efficient security algorithm for VoIP applications running on mobile devices (SecVoIP). The algorithm resolves several weaknesses available in current algorithms while maintaining an appropriate security level. Several experiments were conducted and the results showed significant improvement in processing time, CPU cycles, and consumed energy as compared to SRTP, one of the most widely used security protocols for VoIP. Moreover, we present the results of extensive experimental work that demonstrates that known plaintext attacks against audio streams are not feasible.

Enhanced Multi-rate Multicast for Optimal User Satisfaction and Efficient Bandwidth Utilization

In this paper, we present an enhanced algorithm for a multi-resolution multicast scheme. The enhancement aims at maintaining a perfect 100% satisfaction rate for all the nodes in the network by offering them their requested quality of service. In addition to the multi-resolution multicast, we use simple unicast to the unsatisfied nodes in order to provide the missing layers. The enhancement increases the satisfaction rate from 70% for multicast to the ideal target of 100%, while saving 50% in terms of traffic volume when compared to an all-unicast transmission. A trade-off is also shown between satisfaction rate and the increase in traffic volume.

Energy-efficient HEVC using stochastic processing

One of the serious challenges facing smartphones and other portable devices is battery life. We aim at improving the energy efficiency of multimedia applications for improved battery life while maintaining an acceptable level of user experience. In this paper, we first present the full profiling of the HEVC encoder. Stochastic processing is then applied to the motion estimation stage of HEVC in order to reduce its energy consumption. Finally, video snapshots are displayed along with the relative percentages in energy savings. Savings up to 13% were reached while maintaining a very good quality according to the video quality metric VQM.

Modeling Attacks on MANET Using an Incomplete Repeated Bayesian Game

Nowadays, individuals, as well as corporations, use the internet on a daily basis to send and receive emails, browse the internet, perform financial transactions, etc…However, this dependence comes with the huge risk of communicating over a network that has been compromised. Network administrators are constantly faced with new and improved security attacks that might result in significant losses. Only recently have game theoretic approaches found their way into the area of network security. In this paper we discuss the different types of security attacks while focusing on a general attack that resembles attacks on mobile ad hoc networks or MANETs. We model the attack as a Bayesian repetitive game with incomplete information between a malicious node and a normal node. We study the Nash equilibria of the game, and describe several punishment strategies. Finally, we show the results of our simulations where an equilibrium state is always reached and the average profit of normal nodes always significantly exceeds that of malicious nodes.

Energy analysis of HEVC compression with stochastic processing

One of the serious challenges facing smartphones and other portable devices is battery life. We aim at improving the energy efficiency of multimedia applications for improved battery life while maintaining an acceptable level of user experience. In this paper, we first describe our work in the area of stochastic video compression using the HEVC encoder. We then show the results in energy savings in the encoding stage as a function of an imposed error in addition and subtraction operations. We also present the relationship between the increased bit error rate and the output size of the video. We explain the effect that this increase in file size has on the transmission energy as well as the overall number of instructions, and in particular, the memory read/write operations. We demonstrate the full energy analysis of the entire encoding process using the HEVC encoder and show the quantitative results in every stage.

CrowdApp: Crowdsourcing for application rating

One of the main concerns for application developers is user satisfaction. Before installing any application from an app market, users first look at the app rating and the number of times it was downloaded. However, ratings are not made by experts, are subjective, and require user involvement; therefore, a large number of reviews are needed before the ratings become statistically reliable. One way to obtain user input transparently is to collect data from devices through crowdsourcing. In this work, we present CrowdApp, a crowdsourcing-based application that continuously runs in the background and collects data from the device without any active user participation (transparent crowdsourcing). Then it computes a score for every app installed on the device. The application was tested on Android. Our results show that there is a high correlation (> 0.8) between CrowdApp scores and Google Play scores. More importantly, CrowdApp scores also agreed with subjective ratings by the users themselves at the end of the experiment period.

Playing with Sybil

Recommender systems have become quite popular recently. However, such systems are vulnerable to several types of attacks that target user ratings. One such attack is the Sybil attack where an entity masquerades as several identities with the intention of diverting user ratings. In this work, we propose classical and spatial evolutionary game theory as possible solutions to the Sybil attack in recommender systems. We model the attack using the two techniques and use replicator dynamics in the classical model to solve for evolutionary stable strategies. Results from both models agree that under conditions that are easily achievable by a system administrator, the probability of an attack strategy drops to zero implying degraded fitness for Sybils that eventually die out.