Khurum Nazir Junejo

Automatic Personalized Spam Filtering through Significant Word Modeling

Many real life situations can be modeled as Prisoners dilemma. There are various strategies in the literature. However, few of which match the design objectives of an intelligent agent - being reactive and pro-active. In this paper, we incorporate risk attitude and reputation into infinitely repeated games. In this way, we find that the original game matrix can be transformed to a new matrix, which has a kind of cooperative equilibrium. We use the proposed concepts to analyze the Iterated Prisoners dilemma. Simulation also shows that agents, which consider risk attitude and reputation in the decision-making process, have improved performance and are reactive as well as pro-active.

Behaviour-Based Attack Detection and Classification in Cyber Physical Systems Using Machine Learning

Cyber-physical systems (CPS) are often network integrated to enable remote management, monitoring, and reporting. Such integration has made them vulnerable to cyber attacks originating from an untrusted network (e.g., the internet). Once an attacker breaches the network security, he could corrupt operations of the system in question, which may in turn lead to catastrophes. Hence there is a critical need to detect intrusions into mission-critical CPS. Signature based detection may not work well for CPS, whose complexity may preclude any succinct signatures that we will need. Specification based detection requires accurate definitions of system behaviour that similarly can be hard to obtain, due to the CPSs complexity and dynamics, as well as inaccuracies and incompleteness of design documents or operation manuals. Formal models, to be tractable, are often oversimplified, in which case they will not support effective detection. In this paper, we study a behaviour-based machine learning (ML) approach for the intrusion detection. Whereas prior unsupervised ML methods have suffered from high missed detection or false-positive rates, we use a high-fidelity CPS testbed, which replicates all main physical and control components of a modern water treatment facility, to generate systematic training data for a supervised method. The method does not only detect the occurrence of a cyber attack at the physical process layer, but it also identifies the specific type of the attack. Its detection is fast and robust to noise. Furthermore, its adaptive system model can learn quickly to match dynamics of the CPS and its operating environment. It exhibits a low false positive (FP) rate, yet high precision and recall.

Terms-based discriminative information space for robust text classification

With the popularity of Web 2.0, there has been a phenomenal increase in the utility of text classification in applications like document filtering and sentiment categorization. Many of these applications demand that the classification method be efficient and robust, yet produce accurate categorizations by using the terms in the documents only. In this paper, we propose a novel and efficient method using terms-based discriminative information space for robust text classification. Terms in the documents are assigned weights according to the discrimination information they provide for one category over the others. These weights also serve to partition the terms into category sets. A linear opinion pool is adopted for combining the discrimination information provided by each set of terms to yield a feature space (discriminative information space) having dimensions equal to the number of classes. Subsequently, a discriminant function is learned to categorize the documents in the feature space. This classification methodology relies upon corpus information only, and is robust to distribution shifts and noise. We develop theoretical parallels of our methodology with generative, discriminative, and hybrid classifiers. We evaluate our methodology extensively with five different discriminative term weighting schemes on six data sets from different application areas. We give a side-by-side comparison with four well-known text classification techniques. The results show that our methodology consistently outperforms the rest, especially when there is a distribution shift from training to test sets. Moreover, our methodology is simple and effective for different application domains and training set sizes. It is also fast with a small and tunable memory footprint.

Age estimation from facial images using biometric ratios and wrinkle analysis

A great deal of information can be inferred from human facial images, such as emotions, gender, race, and age. Recently age estimation has developed a variety of applications including, internet access control, underage prevention of cigarette and alcohol machines, and many more. This paper presents a method to classify facial images into 3 evenly distributed age groups. Biometric ratios and wrinkle analysis are used to define features of faces. Three different classifying algorithms have been used for prediction. Testing is done using hold-out approach. A sufficiently large database is used to validate the credibility of results.

Predicting treatment outcome of drug-susceptible tuberculosis patients using machine-learning models

ABSTRACT Tuberculosis (TB) is a deadly contagious disease and a serious global health problem. It is curable but due to its lengthy treatment process, a patient is likely to leave the treatment incomplete, leading to a more lethal, drug resistant form of disease. The World Health Organization (WHO) propagates Directly Observed Therapy Short-course (DOTS) as an effective way to stop the spread of TB in communities with a high burden. But DOTS also adds a significant burden on the financial feasibility of the program. We aim to facilitate TB programs by predicting the outcome of the treatment of a particular patient at the start of treatment so that their health workers can be utilized in a targeted and cost-effective way. The problem was modeled as a classification problem, and the outcome of treatment was predicted using state-of-art implementations of 3 machine learning algorithms. 4213 patients were evaluated, out of which 64.37% completed their treatment. Results were evaluated using 4 performance measures; accuracy, precision, sensitivity, and specificity. The models offer an improvement of more than 12% accuracy over the baseline prediction. Empirical results also revealed some insights to improve TB programs. Overall, our proposed methodology will may help teams running TB programs manage their human resources more effectively, thus saving more lives.

Integrating Design and Data Centric Approaches to Generate Invariants for Distributed Attack Detection

Process anomaly is used for detecting cyber-physical attacks on critical infrastructure such as plants for water treatment and electric power generation. Identification of process anomaly is possible using rules that govern the physical and chemical behavior of the process within a plant. These rules, often referred to as invariants, can be derived either directly from plant design or from the data generated in an operational. However, for operational legacy plants, one might consider a data-centric approach for the derivation of invariants. The study reported here is a comparison of design-centric and data-centric approaches to derive process invariants. The study was conducted using the design of, and the data generated from, an operational water treatment plant. The outcome of the study supports the conjecture that neither approach is adequate in itself, and hence, the two ought to be integrated.

Harnessing English Sentiment Lexicons for Polarity Detection in Urdu Tweets: A Baseline Approach

It is human instinct to express emotions, and with increasing use of social media, it is more often being expressed through text messages than ever before. The emotions and sentiments encoded in these short text messages are of keen interest to various marketing and advertising agencies. Thus, various lexicons and algorithms have been devised for English, and French language to extract these hidden sentiments. On the other hand, Urdu (or Hindi) the third widely-spoken language in the world [1], lacks any such sentiment lexicons or algorithms. Instead of starting from scratch, we make use of the existing English sentiment lexicons to develop the first sentiment lexicon for Urdu. This lexicon will serve as a baseline for future lexicons developed through more intimate knowledge of Urdu language. Furthermore, we compare its performance with various machine learning (ML) approaches. We also make public the labeled dataset developed by us for Urdu sentiment analysis. We hope that this lexicon and dataset will serve as a benchmark for evaluation of future lexicons and ML approaches for the Urdu language.

Distribution shift resilient discrimination information space for SVM classification

There has been a phenomenal increase in the utility of text classification (TC) in applications like targeted advertisement and sentiment analysis. Most applications demand that the model be efficient and robust, yet produce accurate categorizations. This is quite challenging as their is a dearth of labelled training data because it requires assigning labels after reading the whole document. Secondly, the people labelling the documents may not agree on a particular categorization. Therefore, pre-labelled data from a different source, domain, or a user may be used to augment (or replace) the train data. This results in the difference of distribution between the train and test sets. Additionally, with time, the distribution of the test set may change as well. This change in distribution between the train and test sets violates the inductive inference hypothesis that is underlying any machine learning (ML) prediction model, with some ML models being more sensitive towards this phenomenon than others. The performance of support vector machines (SVM) (which is one of the most successful classifiers for TC) degrades drastically in such scenarios. Therefore, in this paper we propose a novel and efficient method that uses terms-based discriminative information space to train SVM for scenarios where distribution shift exists between train and test sets. Our results on eight different train and test pairs from four different domains suggest that the performance gain achieved by SVM trained in the discriminative information space is significantly greater than the performance of SVM trained on the input feature space. Moreover, the methodology is simple, effective, and fast with a small and tunable memory footprint.

Predicting Future Gold Rates using Machine Learning Approach

Historically, gold was used for supporting trade transactions around the world besides other modes of payment. Various states maintained and enhanced their gold reserves and were recognized as wealthy and progressive states. In present times, precious metals like gold are held with central banks of all countries to guarantee re-payment of foreign debts, and also to control inflation. Moreover, it also reflects the financial strength of the country. Besides government agencies, various multi-national companies and individuals have also invested in gold reserves. In traditional events of Asian countries, gold is also presented as gifts/souvenirs and in marriages, gold ornaments are presented as Dowry in India, Pakistan and other countries. In addition to the demand and supply of the commodity in the market, the performance of the world’s leading economies also strongly influences gold rates. We predict future gold rates based on 22 market variables using machine learning techniques. Results show that we can predict the daily gold rates very accurately. Our prediction models will be beneficial for investors, and central banks to decide when to invest in this commodity.

Dynamic gesture recognition using machine learning techniques and factors affecting its accuracy

Kinect, a motion sensing input device for gaming consoles has been successfully utilized for video games, and rehabilitation of paralyzed patients. We use this device to make learning a fun activity for children. Children learn to draw shapes by moving their hands in front of the Kinect device. We automatically recognize and classify their dynamic hand gestures into predefined shapes, namely; rectangles, triangles, and circles. To decrease over fitting and the cost of generating sample shapes a novel feature engineering approach is also proposed that increases the performance by more than 11%. We used three different machine learning algorithms and successfully classified the shapes with an accuracy of more than 97%.