Fast-flux Domains Research Articles

DGA domain embedding with deep metric learning

Abstract Botnets currently use domain-generation algorithms to produce fast-flux domains that enable them to evade detection. Accurately categorizing these botnet domains is crucial to develop cybersecurity solutions against botnet threats. However, existing methods, requiring labeled data, are ineffective against new botnets. To address this issue, we propose Domain2Vec, a metric learning-based approach that can explore new botnets. Domain2Vec integrates a framework of metric learning, which uses individual domains from known botnets for categorization of unknown botnet domains. The training involves an attention-based encoder, and it includes a constraint to ensure that samples with the same labels are closer in the embedding space. The categorization uses the encoder to project domain names into appropriate representations (numerical vectors), even for domains from new botnets. Finally, Domain2Vec uses numerical vectors to explore botnets. Experiments showed that Domain2Vec performs well on domain retrieval and clustering tasks without labeled data, outperforming the state of the art by 13% and 100%, respectively. Real-world tests demonstrate that Domain2Vec can effectively identify unreported malicious domains and monitor botnet activities.

Botnet Attack Detection Using A Hybrid Supervised Fast-Flux Killer System

A Fast Flux Service Network (FFSN) domain name system method is a technique used on botnet that bot herders used to support malicious botnet actions to rapidly change the domain name IP addresses and to increase the life of malicious servers. While several methods for the detection of FFSN domains are suggested, they are still suffering from relatively low accuracy with the zero-day domain in particular. Throughout the current research, a system that’s deemed new is proposed. The latter system is called (the Fast Flux Killer System) and is abbreviated as (FFKS)). It allows one to have the FF-Domains “zero-day”, via a deployment built on (ADeSNN). It is a hybrid, which consists of two stages. The online phase according to the learning outcomes from the offline phase works on detecting the zero-day domains while the offline phase helps in enhancing the classification performance of the system in the online phase. This system will be compared to a previously published work that was based on a supervised detection method using the same ADeSNN algorithm to have the FFSNs domains detected, also to show better performance in detecting malicious domains. A public data set for the impacts of the hybrid ADeSNN algorithm is employed in the experiment. When hybrid ADeSNN was used over the supervised one, the experiments showed better accuracy. The detection of zero-day fast-flux domains is highly accurate (99.54%) in a mode considered as an online one.

PASSVM: A highly accurate fast flux detection system

Fast Flux service networks (FFSNs) are used by adversaries to provide high availability to malicious servers while keeping them hidden from direct access. In these networks, a large number of botnet machines work as proxies to relay the traffic between end users and a malicious mothership server which is controlled by an adversary. Various mechanisms have been proposed for detecting FFSNs. However, most of these mechanisms depend on collecting a large amount of DNS traffic traces and require a considerable amount of time to identify fast flux domains. In this paper, we propose an efficient AI-based online fast flux detection system that performs highly accurate and extremely fast detection of fast flux domains. The proposed system, called PASSVM, is based on features that are associated with DNS response messages of a given domain name. The approach relies on features that are stored in local databases, in addition to features that are extracted from the response DNS messages. The information in the databases are obtained from Censys search engine and an IP Geolocation service. PASSVM is evaluated using three types of supervised machine learning algorithms which are: Multilayer Perceptron (MLP), Radial Basis Function Network (RBF), and Support Vector Machines (SVM). Results show that SVM with RBF kernel outperformed the other two methods with an accuracy of 99.557% and a detection time of less than 18 ms.

A Survey of Fast Flux Botnet Detection With Fast Flux Cloud Computing

A botnet refers to a set of compromised machines controlled distantly by an attacker. Botnets are considered the basis of numerous security threats around the world. Command and control (C&C) servers are the backbone of botnet communications, in which bots send a report to the botmaster, and the latter sends attack orders to those bots. Botnets are also categorized according to their C&C protocols, such as internet relay chat (IRC) and peer-to-peer (P2P) botnets. A domain name system (DNS) method known as fast-flux is used by bot herders to cover malicious botnet activities and increase the lifetime of malicious servers by quickly changing the IP addresses of the domain names over time. Several methods have been suggested to detect fast-flux domains. However, these methods achieve low detection accuracy, especially for zero-day domains. They also entail a significantly long detection time and consume high memory storage. In this survey, we present an overview of the various techniques used to detect fast-flux domains according to solution scopes, namely, host-based, router-based, DNS-based, and cloud computing techniques. This survey provides an understanding of the problem, its current solution space, and the future research directions expected.

TempR: Application of Stricture Dependent Intelligent Classifier for Fast Flux Domain Detection

Fast-flux service networks (FFSN) helps the cyber-criminals to hide the servers used for malicious activities behind a wall of proxies (bots). It provides the reliability and detection evasion to a malicious server. FFSN use a large pool of IP addresses for proxies. Detection of FFSN is difficult as few benign technologies like Content distribution networks and round robin DNS have similar working characteristics. Many approaches have been proposed to detect FFSN and fast flux domains. However, due to dynamic behavior of FFSN, these techniques suffer from a significant number of false positives. In this paper, we present a Temporal and Real time detections based approach (TempR) to detect fast flux domains. The features of fast flux domains and benign domains have been collected and classified using intelligent classifiers. Our technique illustrates 96.99% detection accuracy with the recent behavior of fast flux domains.

Detect Fast-Flux Domains Through Response Time Differences

A fast-flux service network (FFSN) uses dynamic DNS to map a dynamic domain, called fast-flux domain (FF domain), to various IP addresses and uses flux bots to redirect network traffic. Due to its powerful capability to conceal the hosts hidden behind the flux bots, FFSNs are widely adopted by attackers to cover various scams. Although diverse promising solutions have been proposed to detect FF domains, they face the same problem-different countermeasures could be used to bypass their detection. Hence, it becomes a critical issue to develop a new detection solution. According to our survey, unlike normal network services that use dynamic DNS to balance the workloads of their hosts, FFSNs utilize dynamic DNS to hide important bots. As a result, the response time of subsequent requests to an FF domain becomes more fluctuating. Based on the response time differences, this paper develops a new metric, Fast-Flux Score (FF-Score), to detect FF domains. Our system, called fast-flux domain detector (FFDD), is used on a computer that could be an end host or an IDS. A user with a set of unknown URLs, which may be obtained from spam or social networks, can simply determine whether they are benign domains or fast-flux ones using FFDD. Experimental results show that FFDD can accurately detect FF domains with only a 0.3% false positive rate and a 2% false negative rate. It takes less than 20 min for FFDD to determine whether a domain is an FF domain. In addition, FFDD is a lightweight stand-alone system; hence, it does not require special support from an ISP or any other network service.

Hybrid detection and tracking of fast-flux botnet on domain name system traffic

Fast-flux is a Domain Name System (DNS) technique used by botnets to organise compromised hosts into a high-availability, load-balancing network that is similar to Content Delivery Networks (CDNs). Fast-Flux Service Networks (FFSNs) are usually used as proxies of phishing websites and malwares, and hide upstream servers that host actual content. In this paper, by analysing recursive DNS traffic, we develop a fast-flux domain detection method which combines both real-time detection and long-term monitoring. Experimental results demonstrate that our solution can achieve significantly higher detection accuracy values than previous flux-score based algorithms, and is lightweight in terms of resource consumption. We evaluate the performance of the proposed fast-flux detection and tracking solution during a 180-day period of deployment on our university's DNS servers. Based on the tracking results, we successfully identify the changes in the distribution of FFSN and their roles in recent Internet attacks.

Detecting Malicious Fast Flux Domains

Fast-flux service networks (FFSN) are new emerging phenomenon in the internet. Fast-flux networks use proxy networks of compromised machines to redirect and host scam service to achieve high availability. Such technique helps scam websites to avoid being traced and taken down by security professionals. In this paper, we use alternative decision tree algorithm to identify presence of fast-flux domains by analyzing only one address record (A-record) of DNS lookup, which achieves fast detection.

Collaborative Detection of Fast Flux Phishing Domains

Phishing is a significant security threat to users of Internet services. Nowadays, phishing has become more resilient to detection and trace-back with the invention of Fast Flux (FF) service networks. We propose two approaches to correlate evidence from multiple DNS servers and multiple suspect FF domains. Real-world experiments show that our correlation approaches speed-up FF domain detection, based on an analytical model that we propose to quantify the number of DNS queries needed to confirm a FF domain. We also show how our correlation scheme can be implemented on a large scale by using a decentralized publish-subscribe correlation model called LarSID, which is more scalable than a fully centralized architecture.