The integrity of network communications is constantly being challenged by more sophisticated intrusion techniques. Attackers are shifting to stealthier and more complex forms of attacks in an attempt to bypass known mitigation strategies. Also, many detection methods for popular network attacks have been developed using outdated or non-representative attack data. To effectively develop modern detection methodologies, there exists a need to acquire data that can fully encompass the behaviors of persistent and emerging threats. When collecting modern day network traffic for intrusion detection, substantial amounts of traffic can be collected, much of which consists of relatively few attack instances as compared to normal traffic. This skewed distribution between normal and attack data can lead to high levels of class imbalance. Machine learning techniques can be used to aid in attack detection, but large levels of imbalance between normal (majority) and attack (minority) instances can lead to inaccurate detection results. / Includes bibliography. / Dissertation (Ph.D.)--Florida Atlantic University, 2019. / FAU Electronic Theses and Dissertations Collection

<div> <div> <div> <p>Nowadays, the pace of coordinated cyber security crimes has become drastically more rapid, and network attacks have become more advanced and diversified. The explosive growth of network security threats poses serious challenges for building secure Cyber-based Systems (CBS). Existing studies have addressed a breadth of challenges related to detecting network attacks. However, there is still a lack of studies on the detection of sophisticated Multi-stage Attacks (MSAs). </p> <p>The objective of this dissertation is to address the challenges of modeling and detecting sophisticated network attacks, such as multiple interleaved MSAs. We present the interleaving concept and investigate how interleaving multiple MSAs can deceive intrusion detection systems. Using one of the important statistical machine learning (ML) techniques, Hidden Markov Models (HMM), we develop three architectures that take into account the stealth nature of the interleaving attacks, and that can detect and track the progress of these attacks. These architectures deploy a set of HMM templates of known attacks and exhibit varying performance and complexity. </p> <p>For performance evaluation, various metrics are proposed which include (1) attack risk probability, (2) detection error rate, and (3) the number of correctly detected stages. Extensive simulation experiments are conducted to demonstrate the efficacy of the proposed architecture in the presence of multiple multi-stage attack scenarios, and in the presence of false alerts with various rates. </p> </div> </div> </div>

Thesis (MTech (Information Technology))--Cape Peninsula University of Technology, 2018. / With the constant evolution of information systems, companies have to acclimatise to the vast increase of data flowing through their networks. Business processes rely heavily on information technology and operate within a framework of little to no space for interruptions. Cyber attacks aimed at interrupting business operations, false intrusion detections and leaked information burden companies with large monetary and reputational costs. Intrusion detection systems analyse network traffic to identify suspicious patterns that intent to compromise the system. Classifiers (algorithms) are used to classify the data within different categories e.g. malicious or normal network traffic. Recent surveys within intrusion detection highlight the need for improved detection techniques and warrant further experimentation for improvement. This experimental research project focuses on implementing swarm intelligence techniques within the intrusion detection domain. The Ant Tree Miner algorithm induces decision trees by using ant colony optimisation techniques. The Ant Tree Miner poses high accuracy with efficient results. However, limited research has been performed on this classifier in other domains such as intrusion detection. The research provides the intrusion detection domain with a new algorithm that improves upon results of decision trees and ant colony optimisation techniques when applied to the domain. The research has led to valuable insights into the Ant Tree Miner classifier within a previously unknown domain and created an intrusion detection benchmark for future researchers.

Given the continuing advancement of networking applications and our increased dependence upon software-based systems, there is a pressing need to develop improved security techniques for defending modern information technology (IT) systems from malicious cyber-attacks. Indeed, anyone can be impacted by such activities, including individuals, corporations, and governments. Furthermore, the sustained expansion of the network user base and its associated set of applications is also introducing additional vulnerabilities which can lead to criminal breaches and loss of critical data. As a result, the broader cybersecurity problem area has emerged as a significant concern, with many solution strategies being proposed for both intrusion detection and prevention. Now in general, the cybersecurity dilemma can be treated as a conflict-resolution setup entailing a security system and minimum of two decision agents with competing goals (e.g., the attacker and the defender). Namely, on the one hand, the defender is focused on guaranteeing that the system operates at or above an adequate (specified) level. Conversely, the attacker is focused on trying to interrupt or corrupt the system’s operation. In light of the above, this dissertation introduces novel methodologies to build appropriate strategies for system administrators (defenders). In particular, detailed mathematical models of security systems are developed to analyze overall performance and predict the likely behavior of the key decision makers influencing the protection structure. The initial objective here is to create a reliable intrusion detection mechanism to help identify malicious attacks at a very early stage, i.e., in order to minimize potentially critical consequences and damage to system privacy and stability. Furthermore, another key objective is also to develop effective intrusion prevention (response) mechanisms. Along these lines, a machine learning based solution framework is developed consisting of two modules. Specifically, the first module prepares the system for analysis and detects whether or not there is a cyber-attack. Meanwhile, the second module analyzes the type of the breach and formulates an adequate response. Namely, a decision agent is used in the latter module to investigate the environment and make appropriate decisions in the case of uncertainty. This agent starts by conducting its analysis in a completely unknown milieu but continually learns to adjust its decision making based upon the provided feedback. The overall system is designed to operate in an automated manner without any intervention from administrators or other cybersecurity personnel. Human input is essentially only required to modify some key model (system) parameters and settings. Overall, the framework developed in this dissertation provides a solid foundation from which to develop improved threat detection and protection mechanisms for static setups, with further extensibility for handling streaming data.

<p>With an increasingly digitally connected society comes complexity, uncertainty, and risk. Network monitoring, incident management, and digital forensics is of increasing importance with the escalation of cybercrime and other network supported serious crimes. New laws and regulations governing electronic communications, cybercrime, and data retention are being proposed, continuously requiring new methods and tools.</p><p>This thesis introduces a novel approach to real-time network risk assessment based on hidden Markov models to represent the likelihood of transitions between security states. The method measures risk as a composition of individual hosts, providing a precise, fine-grained model for assessing risk and providing decision support for incident response. The approach has been integrated with an existing framework for distributed, large-scale intrusion detection, and the results of the risk assessment are applied to prioritize the alerts produced by the intrusion detection sensors. Using this implementation, the approach is evaluated on both simulated and real-world data.</p><p>Network monitoring can encompass large networks and process enormous amounts of data, and the practice and its ubiquity can represent a great threat to the privacy and confidentiality of network users. Existing measures for anonymization and pseudonymization are analyzed with respect to the trade-off of performing meaningful data analysis while protecting the identities of the users. The results demonstrate that most existing solutions for pseudonymization are vulnerable to a range of attacks. As a solution, some remedies for strengthening the schemes are proposed, and a method for unlinkable transaction pseudonyms is considered.</p><p>Finally, a novel method for performing digital forensic reconstructions in a virtual security testbed is proposed. Based on a hypothesis of the security incident in question, the testbed is configured with the appropriate operating systems, services, and exploits. Attacks are formulated as event chains and replayed on the testbed. The effects of each event are analyzed in order to support or refute the hypothesis. The purpose of the approach is to facilitate reconstruction experiments in digital forensics. Two examples are given to demonstrate the approach; one overview example based on the Trojan defense and one detailed example of a multi-step attack. Although a reconstruction can neither prove a hypothesis with absolute certainty, nor exclude the correctness of other hypotheses, a standardized environment combined with event reconstruction and testing can lend credibility to an investigation and can be a valuable asset in court.</p>

<p>Traditionally, firewalls and access control have been the most important components used in order to secure servers, hosts and computer networks. Today, intrusion detection systems (IDSs) are gaining attention and the usage of these systems is increasing. This thesis covers commercial IDSs and the future direction of these systems. A model and taxonomy for IDSs and the technologies behind intrusion detection is presented. </p><p>Today, many problems exist that cripple the usage of intrusion detection systems. The decreasing confidence in the alerts generated by IDSs is directly related to serious problems like false positives. By studying IDS technologies and analyzing interviews conducted with security departments at Swedish banks, this thesis identifies the major problems within IDSs today. The identified problems, together with recent IDS research reports published at the RAID 2002 symposium, are used to recommend the future direction of commercial intrusion detection systems.</p>

<p>Critical networks require defence in depth incorporating many different security technologies including intrusion detection. One important intrusion detection approach is called anomaly detection where normal (good) behaviour of users of the protected system is modelled, often using machine learning or data mining techniques. During detection new data is matched against the normality model, and deviations are marked as anomalies. Since no knowledge of attacks is needed to train the normality model, anomaly detection may detect previously unknown attacks.</p><p>In this thesis we present ADWICE (Anomaly Detection With fast Incremental Clustering) and evaluate it in IP networks. ADWICE has the following properties:</p><p>(i) Adaptation - Rather than making use of extensive periodic retraining sessions on stored off-line data to handle changes, ADWICE is fully incremental making very flexible on-line training of the model possible without destroying what is already learnt. When subsets of the model are not useful anymore, those clusters can be forgotten.</p><p>(ii) Performance - ADWICE is linear in the number of input data thereby heavily reducing training time compared to alternative clustering algorithms. Training time as well as detection time is further reduced by the use of an integrated search-index.</p><p>(iii) Scalability - Rather than keeping all data in memory, only compact cluster summaries are used. The linear time complexity also improves scalability of training.</p><p>We have implemented ADWICE and integrated the algorithm in a software agent. The agent is a part of the Safeguard agent architecture, developed to perform network monitoring, intrusion detection and correlation as well as recovery. We have also applied ADWICE to publicly available network data to compare our approach to related works with similar approaches. The evaluation resulted in a high detection rate at reasonable false positives rate.</p> / Report code: LiU-Tek-Lic-2006:12.

<p>Many of the attacks on computing clusters and grids have been performed by using stolen authentication passwords and unprotected SSH keys, therefore there is a need for a system that can detect intruders masquerading as ordinary users. Our assumption is that an attacker behaves significantly different compared to an ordinary user. Previous work in this area is for example statistical analysis of process accounting using Support Vector Machines. We can formalize this into a classification problem that we will solve with Self-organizing maps. The proposed system will work in a tier model that uses process accounting and SSH log messages as data sources.</p>

The increasing number of network security related incidents has made it necessary for the organizations to actively protect their sensitive data with network intrusion detection systems (IDSs). Detecting intrusion in a distributed network from outside network segment as well as from inside is a difficult problem. IDSs are expected to analyze a large volume of data while not placing a significant added load on the monitoring systems and networks. This requires good data mining strategies which take less time and give accurate results. In this study, a novel hybrid layered multiagent-based intrusion detection system is created, particularly with the support of a multi-class supervised classification technique. In agent-based IDS, there is no central control and therefore no central point of failure. Agents can detect and take predefined actions against malicious activities, which can be detected with the help of data mining techniques. The proposed IDS shows superior performance compared to central sniffing IDS techniques, and saves network resources compared to other distributed IDSs with mobile agents that activate too many sniffers causing bottlenecks in the network. This is one of the major motivations to use a distributed model based on a multiagent platform along with a supervised classification technique. Applying multiagent technology to the management of network security is a challenging task since it requires the management on different time instances and has many interactions. To facilitate information exchange between different agents in the proposed hybrid layered multiagent architecture, a low cost and low response time agent communication protocol is developed to tackle the issues typically associated with a distributed multiagent system, such as poor system performance, excessive processing power requirement, and long delays. The bandwidth and response time performance of the proposed end-to-end system is investigated through the simulation of the proposed agent communication protocol on our private LAN testbed called Hierarchical Agent Network for Intrusion Detection Systems (HAN-IDS). The simulation results show that this system is efficient and extensible since it consumes negligible bandwidth with low cost and low response time on the network.

With an increasingly digitally connected society comes complexity, uncertainty, and risk. Network monitoring, incident management, and digital forensics is of increasing importance with the escalation of cybercrime and other network supported serious crimes. New laws and regulations governing electronic communications, cybercrime, and data retention are being proposed, continuously requiring new methods and tools. This thesis introduces a novel approach to real-time network risk assessment based on hidden Markov models to represent the likelihood of transitions between security states. The method measures risk as a composition of individual hosts, providing a precise, fine-grained model for assessing risk and providing decision support for incident response. The approach has been integrated with an existing framework for distributed, large-scale intrusion detection, and the results of the risk assessment are applied to prioritize the alerts produced by the intrusion detection sensors. Using this implementation, the approach is evaluated on both simulated and real-world data. Network monitoring can encompass large networks and process enormous amounts of data, and the practice and its ubiquity can represent a great threat to the privacy and confidentiality of network users. Existing measures for anonymization and pseudonymization are analyzed with respect to the trade-off of performing meaningful data analysis while protecting the identities of the users. The results demonstrate that most existing solutions for pseudonymization are vulnerable to a range of attacks. As a solution, some remedies for strengthening the schemes are proposed, and a method for unlinkable transaction pseudonyms is considered. Finally, a novel method for performing digital forensic reconstructions in a virtual security testbed is proposed. Based on a hypothesis of the security incident in question, the testbed is configured with the appropriate operating systems, services, and exploits. Attacks are formulated as event chains and replayed on the testbed. The effects of each event are analyzed in order to support or refute the hypothesis. The purpose of the approach is to facilitate reconstruction experiments in digital forensics. Two examples are given to demonstrate the approach; one overview example based on the Trojan defense and one detailed example of a multi-step attack. Although a reconstruction can neither prove a hypothesis with absolute certainty, nor exclude the correctness of other hypotheses, a standardized environment combined with event reconstruction and testing can lend credibility to an investigation and can be a valuable asset in court.