Qatar Foundation Annual Research Conference Proceedings Volume 2018 Issue 3

Capability oriented requirements engineering is an emerging research area where designers are faced with the challenge of analyzing changes in the business domain, capturing user requirements, and developing adequate IT solutions taking into account these changes and answering user needs. In this context, researching the interplay between design-time and run-time requirements with a focus on adaptability is of great importance. Approaches to adaptation in the requirements engineering area consider issues underpinning the awareness of requirements and the evolution of requirements. We are focusing on researching the influence of capability-driven requirements on architectures for adaptable systems to be utilized in smart city operations. We investigate requirements specification, algorithms, and prototypes for smart city operations with a focus on intelligent management of transportation and on validating the proposed approaches. In this framework, we conducted a systematic literature review (SLR) of requirements engineering approaches for adaptive system (REAS). We investigated the modeling methods used, the requirements engineering activities performed, the application domains involved, and the deficiencies that need be tackled (in REAS in general, and in SCOs in particular). We aimed at providing an updated review of the state of the art in order to support researchers in understanding trends in REAS in general, and in SCOs in particular. We also focused on the study of Requirement Traceability Recovery (RTR). RTR is the process of constructing traceability links between requirements and other artifacts. It plays an important role in many parts of the software life-cycle. RTR becomes more important and exigent in the case of systems that change frequently and continually, especially adaptive systems where we need to manage the requirement changes in such systems and analyze their impact. We formulated RTR as a mono and a multi-objective search problem using a classic Genetic Algorithm (GA) and a Non-dominated Sorting-based Genetic Algorithm (NSGA-II) respectively. The mono-objective approach takes as input the software system, a set of requirements and generates as output a set of traces between the artifacts of the system and the requirements introduced in the input. This is done based on the textual similarity between the description of the requirements and the artifacts (name of code elements, documentation, comments, etc.). The multi-objective approach takes into account three objectives, namely, the recency of change, the frequency of change, and the semantic similarity between the description of the requirement and the artifact. To validate the two approaches, we used three different open source projects. The reported results confirmed the effectiveness of the two approaches in correctly generating the traces between the requirements and artifacts with high precision and a recall. A comparison between the two approaches shows that the multi-objective approach is more effective than the mono-objective one. We also proposed an approach aiming at optimizing service composition in service-oriented architectures in terms of security goals and cost using NSGA-II in order to help software engineers to map the optimized service composition to the business process model based on security and cost. To do this, we adapted the DREAD model for security risk assessment by suggesting new categorizations for calculating DREAD factors based on a proposed service structure and service attributes. To validate the proposal, we implemented the YAFA-SOA Optimizer. The evaluation of this optimizer shows that risk severity for the generated service composition is less than 0.5, which matches the validation results obtained from a security expert. We also investigated requirements modeling for an event with a large crowd using the capability-oriented paradigm. The motivation was the need for the design of services that meet the challenges of alignment, agility, and sustainability in relation to dynamically changing enterprise requirements especially in large-scale events such as sports events. We introduced the challenges to stakeholders involved in this process and advocated a capability-oriented approach for successfully addressing these challenges. We also investigated a multi-type, proactive and context-aware recommender system in the environment of smart cities. The recommender system recommends gas stations, restaurants, and attractions, proactively, in an internet of things environment. We used a neural network to do the reasoning and validated the system on 7000 random contexts. The results are promising. We also conducted a user's acceptance survey (on 50 users) that showed satisfaction with the application. We also investigated capturing uncertainty in adaptive intelligent transportation systems, which need to monitor their environment at run-time and adapt their behavior in response to changes in this environment. We modelled an intelligent transportation case study using the KAOS goal model and modelled uncertainty by extending our case study using variability points, and hence having different alternatives to choose from depending on the context at run-time. We handled uncertainty by molding our alternatives using ontologies and reasoning to select the optimal alternative at run-time when uncertainty occurs. We also devised a framework, called Vehicell that exploits 5G mobile communication infrastructures to increase the effectiveness of vehicular communications and enhance the relevant services and applications offered in urban environments. This may help in solving some of the mobility problems, and smooth the way for innovative services to citizens and visitors and improve the overall quality of life.

It has been demonstrated that encrypting confidential data before storing it is not sufficient because data access patterns can leak significant information about the data itself (Goldreich & Ostrovsky, 1996). Oblivious RAM (ORAM) schemes exist in order to protect the access pattern of data in a data-store. Under an ORAM algorithm, a client accesses a data store in such a way that does not reveal which item it is interested in. This is typically accomplished by accessing multiple items each access and periodically reshuffling some, or all, of the data on the data-store. One critical limitation of ORAM techniques is the need to have large storage capacity on the client, which is typically a weak device. In this work, we utilize an ORAM technique that adapts itself to working with clients having very limited storage. A trivial implementation for an oblivious RAM scans the entire memory for each actual memory access. This scheme is called linear ORAM. Goldreich and Ostrovsky (Goldreich & Ostrovsky, 1996) presented two ORAM constructions with a hierarchical layered structure: the first, Square-root ORAM, provides square root access complexity and constant space requirement; the second, Hierarchical ORAM, requires logarithmic space and polylogarithmic access complexity. Square-root ORAM was revisited by (Zahur, et al.) to improve its performance in a multi-party secure computation setting. The work of Shi et al. (Shi, Chan, Stefanov, & Li, 2011) adopted a sequence of binary trees as the underlying structure. (Stefanov, et al., 2013) utilized this concept to build a simple ORAM called path ORAM. In path ORAM, every block (item) of data in the input array A is mapped to a (uniformly) random leaf in a tree (typically a binary tree) on the server. This is done using a position map stored in the client memory. Each node in the tree has exactly Z blocks which are initially dummy blocks. Each data item is stored in a node on the path extending from the leaf to which the data item was mapped, to the root. When a specific item is requested, a position map is used to point out the leaf to which the block is mapped. Then the whole path is read starting from the block's mapped leaf up to the root into a stash. We call this procedure a read path method. The stash is a space which also exists on the client. The block is mapped to a new leaf (uniformly random). The client gets the required block from the stash. We then try to evict the contents of the stash to the same path which we read from starting from the leaf to the root. We call this procedure a write path method. To transfer a block into a node in the tree: - The node that is tested should have enough space. - The node should be on the path to which the tested block in the stash is mapped. If both conditions are met, a block is evicted to the tested node. Clearly, the tree is encrypted and whenever the client reads a path into the stash, all read blocks are decrypted and the requested block is sent to the client. The client encrypts the blocks before writing them back to the path. The security proof of this ORAM type is explained in (Stefanov, et al., 2013). We assume that a position map can be fit in the client's memory. Since it requires O(N) space, that could be a problem. (Stefanov, et al., 2013) mentioned a general idea for a solution that uses another smaller ORAM O1 on the server to store the position map and stores the position map for O1 in the client's memory. We employ a recursive generalized version of this approach. If the position map is still larger than the client's capacity, a smaller ORAM O2 is built to store the position map for O1. We call these additional trees auxiliary trees. We implemented the recursive technique for Path ORAM and studied the effect of the threshold size of position map (and consequently, the number of auxiliary trees) on the performance of path ORAM. We tested our implementation on 1 million items using several threshold sizes of position map. The number of accesses is 10,000 in all tests. Our results show expected negative correlation between the time consumption and the threshold size of the position map. However, the results suggest that unless the increase in the threshold size of position map decreases the number of trees, no significant improvement in performance will be noticed. It is also clear that the initialization process which is the process of building the items» tree and the auxiliary trees and filling the initial values comprises more than 98% of the consumed time. Accordingly, this type of ORAM suits the case of large number of accesses since the server can fulfil client»s request very fast after finishing the initialization process. References Goldreich, O., & Ostrovsky, R. (1996). Software protection and simulation on oblivious rams. Journal of the ACM (JACM), vol. 43, no. 3, pp. 431–473. Shi, E., Chan, T.-H., Stefanov, E., & Li, M. (2011). Oblivious ram with o ((logn) 3) worst-case cost. International Conference on The Theory and Application of Cryptology and Information Security. Springer, pp. 197–214. Stefanov, E., Dijk, M. V., Shi, E., Fletcher, C., Ren, L., Yu, X., et al. (2013). Path oram: an extremely simple oblivious ram protocol. Proceedings of the 2013 ACM SIGSAC conference on Computer & communications security. ACM, pp. 299–310.. Zahur, S., Wang, X., Raykova, M., Gascon, A., Doerner, J., Evans, D., et al. (n.d.). Revisiting square-root oram: Efficient random access in multi-party computation. 2016: Security and Privacy (SP), IEEE Symposium on. IEEE, pp. 218–234.

This study presents a preliminary feasibility investigation of signal propagation and antenna diversity techniques inside the human skin tissues in the frequency range (0.8-1.2)Terahertz (THz) by applying multiple input-single output (MISO) technique. THz application in in-vivo communication has received a great attention for its unique properties as non ionizing characteristics, strong interaction with water content inside the human tissues, and molecular sensitivity [1]. This study helps to evaluate the usage and the performance of MISO system for nanoscale network. The human skin tissue is represented by three main layers: stratum corneum, epidermis, and dermis. The path loss model and the channel characterization inside the human skin was investigated in [2]. The diversity gain (DG) for two different in-vivo channels resulting from the signal propagation between two transmitting antennas, located at the dermis layer, and one receiving antenna, located at epidermis layer, is calculated to evaluate the system performance. Different diversity combining techniques are applied in this study: selection combining (SC), equal-gain combing (EGC), and maximum-ratio combining (MRC). In the simulation setting in CST microwave studio, the distance between the transmitting antennas is fixed; while the effect of the distance between the receiver and the transmitters is analyzed at different frequencies. Although MIMO antenna systems are used in wireless communication to enhance data throughput, from the initial study it is predicted that might be it is not useful in in-vivo nano communication. Results demonstrates that there is a high cross correlation between the two channels. Figure 1 shows the CDF plot for the two channel with different diversity combining techniques used in the study.[1] J. M. Jornet and I. F. Akyildiz, “Channel modeling and capacity analysis for electromagnetic wireless nanonetworks in theterahertz band,” IEEE Transactions on Wireless Communications, vol. 10, no. 10, pp. 3211–3221, 2011.[2] Q. H. Abbasi, H. El Sallabi, N. Chopra, K. Yang, K. A. Qaraqe, and A. Alomainy, “Terahertz channel characterizationinside the human skin for nano-scale body-centric networks,” IEEE Transactions on Terahertz Science and Technology, vol.6, no. 3, pp. 427–434, 2016.

Detection of images or moving objects have been highly worked upon, and has been integrated and used in commercial, residential and industrial environments. But, most of the strategies and techniques have heavy limitations. One of the limitations is due to low computational resources at user level. Other important limitations that need to be tackled are lack of proper data analysis of the measured trained data, dependency on the motion of the objects, inability to differentiate one object from other, and also concern over speed of the object under detection and Illuminacy. Hence, there is a need to draft, apply and recognize new techniques of detection that tackle the existing limitations. In our project we have worked upon a model based on Scalable Object Detection, using Deep Neural Networks to localize and track people, cars, potted plants and 16 others categories in the camera preview in real-time. The large Visual Recognition ImageNet package ‘inception5h’ from google is used. This is a trained model, with images of the respective categories, which is then converted to a graph file using neural networks. The graph nodes are usually huge in number and these are optimized for the use in android. The use of already available trained model is just for the purpose of ease and convenience, nevertheless any set of images can be trained and used in the android application. An important point to note is that training the images will need huge computational speeds and more than one computer supporting GPU. Also a.jar file built with the help of bazel is added to android studio to support the integration of Java and tensorflow. This jar file is the key to getting tensorflow in a mobile device. The jar file is built with the help of openCV, which is a library of programming functions mainly aimed at real-time computer vision. Once this has been integrated, any input to the android application inputed at real-time, is predicted with the help of tiny-yolo (you look only once – a darknet reference network). This application supports multi object detection, which is very useful. All the steps occur simultaneously with great speeds giving remarkable results, detecting all the categories of the trained model within a good illuminacy. The real time detection reference network used also works with an acceptable acceleration of moving objects but is not quite effective with low illumination. The objects are limited to identify only 20 categories but the scope can be broadened with a revised trained model. The 20 categories include «aeroplane»,»bicycle»,»bird»,»boat», «bottle», «bus», «car», «cat», «chair», «cow»,»diningtable»,»dog»,»horse»,»motorbike»,»person», «pottedplant», «sheep»,»sofa»,»train» and «tvmonitor». The application can be used handy in a mobile phone or any other smart device with minimal computational resources ie.., no connectivity to the internet. The application challenges speed and illuminacy. Effective results will help in real-time detection of traffic signs and pedestrians from a moving vehicle. This goes hand in hand with the similar intelligence in cameras which can be used as an artificial eye, and can be used in many areas such as surveillance, Robotics, Traffic, facial recognition, etc.

Spatiotemporal data related to traffic has become common place due to the wide availability of cheap sensors and the rapid deployment of IoT platforms. Yet, this data suffer several challenges related to sparsity, incompleteness, and noise, which makes traffic analytics difficult. In this paper, we investigate the problem of missing data or noisy information in the context of real-time monitoring and forecasting of traffic congestion for road networks. The road network is represented as a directed graph in which nodes are junctions and edges are road segments. We assume that the city has deployed high-fidelity sensors for speed reading in a subset of edges. Our objective is to infer speed readings for the remaining edges in the network as well as missing values to malfunctioning sensors. We propose a tensor representation for the series of road network snapshots, and develop a regularized factorization method to estimate the missing values, while learning the latent factors of the network. The regularizer, which incorporates spatial properties of the road network, improves the quality of the results. The learned factors along with a graph-based temporal dependency are used in an autoregressive algorithm to predict the future state of the road network with long horizon. Extensive numerical experiments with real traffic data from the cities of Doha(Qatar) and Aarhus (Denmark) demonstrate that the proposed approach is appropriate for imputing missing data and predicting traffic state.Main contributions. The main contributions are:We propose a novel temporal regularized tensor factorization framework (TRTF) for high-dimensional traffic data. TRTF provides a principled approach to account for both the spatial structure and the temporal dependencies.We introduce a novel data-driven graph-based autoregressive model, where the weights are learned from the data. Hence, the regularizer can account for both positive and negative correlations.We show that incorporating temporal embeddings into CP-WOPT leads to accurate multi-step forecasting, compared to state of the art matrix factorization based methods.We conduct extensive experiments on real traffic congestion datasets from two different cities and show the superiority of TRTF for both tasks of missing value completion and multi-step forecasting under different experimental settings. For instance,TRTF outperforms LSM-RN by 24% and TRMF by 29%.Conclusion. We present in this paper TRTF, an algorithm for temporal regularized tensor decomposition. We show how the algorithm can be used for several traffic related tasks such as missing value completion and forecasting. The proposed algorithm incorporates both spa-tial and temporal properties into the tensor decomposition procedures such as CP-WOPT, yielding to learning better factors. We also, extend TRTF with an auto-regressive procedure to allow for multi step-ahead forecasting of future values. We compare our method to recently developed algorithms that deal with the same type of problems using regularized matrix factorization,and show that under many circumstances, TRTF does provide better results. This is particularly true in cases where the data suffers from high proportions of missing values, which is common in the traffic context. For instance, TRTF achieves a 20% gain in MAPE score compared to the second best algorithm (CP-WOPT) in completing missing values in the case of extreme sparsity observed in Doha. As future work, we will first focus on adding non-negativity constraints to TRTF, although the highest fraction of negative values generated by our method throughout all the experiments did not exceed 0.7%. Our second focus will be to optimize TRTF training phase in order to increase its scalability to handle large dense tensors, and to implement it on a parallel environment.

In the past decade, Cloud-Computing emerged as a new computing concept with a distributed nature using virtual network and systems. Many businesses rely on this technology to keep their systems running but concerns are rising about security breaches in cloud computing. This work presents a secure approach for storing data on the cloud. the proposed methodology is described as follows: 1) The client who wants to store data on the cloud subscribes with n cloud providers (CPs). 2) A file F to be stored on the cloud is subdivided into n parts, or subfiles: F1, F2,.... Fn. 3) Each part is encrypted with an encryption key Kf. The encrypted parts are denoted by F1*, F2*,…, Fn*. 4) A random permutation vector P(f) is generated. 5) The encrypted parts are stored on the n clouds according to P(F); in other words, due to P(F), F1* could be stored on CP3 for example F2* on CPn, etc. 6) In order to be able to retrieve his files, the client needs to maintain some information related to the distribution of the various parts and to the encryption of the file. Thus, he maintains two tables. 7) The first table contains a hash of the file name H(F_name), and the key Kf. 8) The second table contains a hash of the file name H(F_name), a hash of the file content itself (unencrypted) H(F), a hash of the encrypted file content H(F*), and the permutation vector P(F), encrypted with Kf. 9) The two tables are stored on different servers protected by advanced security measures, and preferably located at different locations. 10) In order to obtain the file, the client enters the file name. Then the system computes the hash value of the name, finds the corresponding entry in Table 1, and obtains the key. Then, the corresponding entry in Table 2 is found. The key, obtained from Table 1, is used to decrypt the permutation vector P(f). Then, the encrypted parts are downloaded from the different cloud providers. Afterwards, they are assembled in the correct order and decrypted. The hash values of the encrypted and unencrypted versions are then computed and compared to their corresponding values stored in Table2 in order to check for the integrity of the file downloaded from the cloud. This approach allows the client to use the same storage space on the cloud: Instead of using a single cloud provider to store a file of size S bits, the client is using n cloud providers, where he stores S/n bits with each cloud provider. Thus, the storage costs of the two methods are comparable. If the client wishes to introduce redundancy in the file, such that he can recover the whole file from j parts instead of n parts, with j< = n, then redundancy can be added to the original file as appropriate. In this case, the storage costs will increase accordingly, but this is an added enhancement that can be used with or without the proposed approach. On the other hand, the overhead due to the proposed approach consists of maintaining two tables containing the information relevant to the file. The storage required to maintain the entry corresponding to each file in these two tables is small compared to a typical file size: in fact, we only need to store a few hash values (of fixed size), along with the encryption key. This seems a reasonable price to pay for a client that has sensitive data that he cannot post unencrypted on the cloud, or even posting it encrypted with a single provider is risky in case a security breach occurs at the provider»s premises.

Redundant Array of Independent Disks (RAID) storage architectures provide protection of digital infrastructure against potential disks failures. For example, RAID-5 and RAID-6 architectures provide protection against one and two disk failures, respectively. Recently, the data generation has significantly increased due to the emergence of new technologies. Thus, the size of storage systems is also growing rapidly to accommodate such large data sizes, which increases the probability for disks failures. This necessitates a new RAID architecture that can tolerate up to three disk failures. RAID architectures implement coding techniques. The code specifies how data is stored among multiple disks and how lost data can be recovered from surviving disks. This abstract introduces a novel coding scheme for new RAID-7 architectures that can tolerate up to three disks failures. The code is an improved version of the existing BP-XOR code and is called “Almost BP-XOR”.There are multiple codes that can be used for RAID-7 architectures. However, [5,2] BP-XOR codes have significantly lower encoding and decoding complexities than most common codes [1]. Regardless of this fact, this code does not achieve the fastest data decoding and reconstruction speeds due to its relatively low efficiency of 0.4. Furthermore, the existence of MDS [6,3] bx6 BP-XOR codes, b>2 (which achieves efficiency of 0.5) is still an open research question. This work proposes [6,3] 2 x 6 Almost BP-XOR codes. These codes largely utilize the simple and fast BP-XOR decoder while achieving an efficiency of 0.5, leading to the fastest recovery from disk failures among other state-of-the-art codes. An algorithm to generate a [6, 3] 2 x 6 Almost BP-XOR code has been developed and an example code is provided in Table 1. The [6, 3] 2 x 6 Almost BP-XOR codes are constructed in a way that any three-column-erasure pattern will result in one of the following two main scenarios. First: At least one of the surviving degree-three encoding symbols contains two known information symbols. This scenario occurs in 70% of three-column erasure cases (i.e. 14 out of the 20 possible cases). The recovery process in such scenario is identical to that of the BP-XOR codes; Knowing any two information symbols in a degree-three encoding symbol is sufficient to know the third information symbol through performing a simple XOR operation. Second: None of the surviving degree-three encoding symbols contains two known information symbols. This scenario occurs in the remaining 30% of three-column erasure cases (i.e., 6 out if the possible 20). The BP-XOR decoder fails in such a scenario. However, due to the construction of the codes, at least one surviving degree-three encoding symbol contains a known information symbol. Thus, knowing one of the reaming two information symbols in such a degree-three encoding symbol will initiate the BP-XOR decoder again.Table 2 shows these erasure patterns along with an expression for one of the missing information symbols. these expressions can be stored in buffers and used whenever the corresponding erasure pattern occurs. Solutions in Table 2 are derived from the inverse of a 6x6 submatrix that results from a generator matrix G by deleting columns from G corresponding to erased code columns. The read complexity of almost BP-XOR codes is 1. On the other hand. The decoding of almost BP-XOR codes require just 6 XOR operations when for a given three-column-erasure pattern BP-XOR decoding succeeds. However, when the BP- XOR decoder fails, it will require up to 15 XOR operations in total. The normalized repairing complexity is 15/6 = 2.5.Experimentally, Fig.1 shows that the proposed Almost BP-XOR codes require the least amount of time to decode and reconstruct erased columns. Thus, it is concluded that the [6, 3] 2x6 almost BP-XOR codes are best suited for RAID-7 system that requires storage efficiency of 0.5. References [1] Y. Wang. “Array BP-XOR codes for reliable cloud storage systems,” in Proc. of the 2013 IEEE International Symposium on Information Theory (ISIT), pp. 326–330, Istanbul, Turkey, July 2013.NoteFigures, Tables, and more details are provided in the complete attached file titled Abstract-ARC18.pdf, (respecting the same word count restriction).

In an mHealth remote patient monitoring scenario, usually control units/data aggregators receive data from the body area network (BAN) sensors then send it to the network or “cloud”. The control unit would have to transmit the measurement data to the home access point (AP) using WiFi for example, or directly to a cellular base station (BS), e.g. using the long-term evolution (LTE) technology, or both (e.g. using multi-homing to transmit over multiple radio access technologies (Multi-RATs). Fast encryption or physical layer security techniques are needed to secure the data. In fact, during normal conditions, monitoring data can be transmitted using best effort transmission. However, when real-time processing detects an emergency situation, the current monitoring data should be transmitted real-time to the appropriate medical personnel in emergency response teams. The proposed approach consists of benefiting of the presence of multi-RATs in order to exchange the secrecy information more efficiently while optimizing the transmission time. It can be summarized as follows (assuming there are two RATs):1) The first step is to determine the proportion of data bits to be transmitted over each RAT in order to minimize the transmission time, given the data rates achievable on each RAT. Denoting the data rates by R1, and R2, and the total number of bits to be transmitted by D =  SUM(D1,D2), where D1 and D2 are the number of bits to be transmitted over RAT1 and RAT2 respectively, then they should be selected such that D1/R1 = D2/R22) Then, the exchange of the secrecy parameters between sender and receiver is done over the two RATs in order to maintain the security of the transmission. To avoid the complexity of public key cryptography, a three way handshake can be used: 2-1) The sender decides to divide the data into n parts, with a fraction n1 sent on RAT1 and a fraction n2 sent on RAT2, according to the ratios determined in 1) above (i.e. the sum of the bits in the n1 parts should be close to D1 bits, and the sum of the n2 parts should be close to D2 bits) 2-2) The sender generates a scrambling vector P(D,n) to scramble the n data parts and transmit them out of order. 2-3) The sender groups the secret information consisting of S =  {n, n1, n2, P(D,n)}, and could add additional information to protect against replay attacks, e.g. timestamp, nonce, etc., and sends this information on the two RATs, encrypted by a different key: K11 on RAT1 and K12 on RAT2. Thus, {S}_K11 is sent on RAT1 and {S}_K12 is sent on RAT2. 2-4) The receiver does not know K11 and K12. Thus, it encrypts the received information with two other keys K21 (over RAT1) and K22 (over RAT2) and sends them back: {{S}_K11,K21} is sent on RAT1 and {{S}_K12,K22} is sent on RAT2. 2-5) The sender decodes the received encrypted vectors using his keys and sends back {S}_K21 on RAT1 and {S}_K22 on RAT2. The secret information is still securely encoded by the receiver's secret keys K21 and K22. 2-6) The receiver can now decrypt the information and obtain S. 3. The two parties can now communicate using the secret scrambling approach provided by S. This information can be changed periodically as needed. For example, if the data is subdivided over 10 parts, with 40% to be sent over LTE and 60% to be sent over WiFi, according to the scrambling vector P(D,n) = {4,1,10,7,3,9,5,2,6}, then parts {4,1,10,7} are sent over LTE and parts {7,3,9,5,2,6} are sent over WiFi. The receiver will sort them out in the correct order.