Yazan Boshmaf, M. Ripeanu, K. Beznosov, E. Santos-Neto
Traditional defense mechanisms for fighting against automated fake accounts in online social networks are victim-agnostic. Even though victims of fake accounts play an important role in the viability of subsequent attacks, there is no work on utilizing this insight to improve the status quo. In this position paper, we take the first step and propose to incorporate predictions about victims of unknown fakes into the workflows of existing defense mechanisms. In particular, we investigated how such an integration could lead to more robust fake account defense mechanisms. We also used real-world datasets from Facebook and Tuenti to evaluate the feasibility of predicting victims of fake accounts using supervised machine learning.
{"title":"Thwarting Fake OSN Accounts by Predicting their Victims","authors":"Yazan Boshmaf, M. Ripeanu, K. Beznosov, E. Santos-Neto","doi":"10.1145/2808769.2808772","DOIUrl":"https://doi.org/10.1145/2808769.2808772","url":null,"abstract":"Traditional defense mechanisms for fighting against automated fake accounts in online social networks are victim-agnostic. Even though victims of fake accounts play an important role in the viability of subsequent attacks, there is no work on utilizing this insight to improve the status quo. In this position paper, we take the first step and propose to incorporate predictions about victims of unknown fakes into the workflows of existing defense mechanisms. In particular, we investigated how such an integration could lead to more robust fake account defense mechanisms. We also used real-world datasets from Facebook and Tuenti to evaluate the feasibility of predicting victims of fake accounts using supervised machine learning.","PeriodicalId":426614,"journal":{"name":"Proceedings of the 8th ACM Workshop on Artificial Intelligence and Security","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125718174","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Fake accounts are a preferred means for malicious users of online social networks to send spam, commit fraud, or otherwise abuse the system. A single malicious actor may create dozens to thousands of fake accounts in order to scale their operation to reach the maximum number of legitimate members. Detecting and taking action on these accounts as quickly as possible is imperative in order to protect legitimate members and maintain the trustworthiness of the network. However, any individual fake account may appear to be legitimate on first inspection, for example by having a real-sounding name or a believable profile. In this work we describe a scalable approach to finding groups of fake accounts registered by the same actor. The main technique is a supervised machine learning pipeline for classifying {em an entire cluster} of accounts as malicious or legitimate. The key features used in the model are statistics on fields of user-generated text such as name, email address, company or university; these include both frequencies of patterns {em within} the cluster (e.g., do all of the emails share a common letter/digit pattern) and comparison of text frequencies across the entire user base (e.g., are all of the names rare?). We apply our framework to analyze account data on LinkedIn grouped by registration IP address and registration date. Our model achieved AUC 0.98 on a held-out test set and AUC 0.95 on out-of-sample testing data. The model has been productionalized and has identified more than 250,000 fake accounts since deployment.
{"title":"Detecting Clusters of Fake Accounts in Online Social Networks","authors":"Cao Xiao, D. Freeman, Theodore Hwa","doi":"10.1145/2808769.2808779","DOIUrl":"https://doi.org/10.1145/2808769.2808779","url":null,"abstract":"Fake accounts are a preferred means for malicious users of online social networks to send spam, commit fraud, or otherwise abuse the system. A single malicious actor may create dozens to thousands of fake accounts in order to scale their operation to reach the maximum number of legitimate members. Detecting and taking action on these accounts as quickly as possible is imperative in order to protect legitimate members and maintain the trustworthiness of the network. However, any individual fake account may appear to be legitimate on first inspection, for example by having a real-sounding name or a believable profile. In this work we describe a scalable approach to finding groups of fake accounts registered by the same actor. The main technique is a supervised machine learning pipeline for classifying {em an entire cluster} of accounts as malicious or legitimate. The key features used in the model are statistics on fields of user-generated text such as name, email address, company or university; these include both frequencies of patterns {em within} the cluster (e.g., do all of the emails share a common letter/digit pattern) and comparison of text frequencies across the entire user base (e.g., are all of the names rare?). We apply our framework to analyze account data on LinkedIn grouped by registration IP address and registration date. Our model achieved AUC 0.98 on a held-out test set and AUC 0.95 on out-of-sample testing data. The model has been productionalized and has identified more than 250,000 fake accounts since deployment.","PeriodicalId":426614,"journal":{"name":"Proceedings of the 8th ACM Workshop on Artificial Intelligence and Security","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125491559","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Alex Kantchelian, Michael Carl Tschantz, Sadia Afroz, Brad Miller, Vaishaal Shankar, Rekha Bachwani, A. Joseph, J. D. Tygar
We examine the problem of aggregating the results of multiple anti-virus (AV) vendors' detectors into a single authoritative ground-truth label for every binary. To do so, we adapt a well-known generative Bayesian model that postulates the existence of a hidden ground truth upon which the AV labels depend. We use training based on Expectation Maximization for this fully unsupervised technique. We evaluate our method using 279,327 distinct binaries from VirusTotal, each of which appeared for the first time between January 2012 and June 2014. Our evaluation shows that our statistical model is consistently more accurate at predicting the future-derived ground truth than all unweighted rules of the form "k out of n" AV detections. In addition, we evaluate the scenario where partial ground truth is available for model building. We train a logistic regression predictor on the partial label information. Our results show that as few as a 100 randomly selected training instances with ground truth are enough to achieve 80% true positive rate for 0.1% false positive rate. In comparison, the best unweighted threshold rule provides only 60% true positive rate at the same false positive rate.
我们研究了将多个反病毒(AV)供应商的检测器的结果聚合为每个二进制文件的单个权威真值标签的问题。为此，我们采用了一个著名的生成贝叶斯模型，该模型假设AV标签所依赖的隐藏基础真理的存在。对于这种完全无监督的技术，我们使用基于期望最大化的训练。我们使用VirusTotal的279,327个不同的二进制文件来评估我们的方法，每个二进制文件都是在2012年1月至2014年6月之间首次出现的。我们的评估表明，我们的统计模型在预测未来衍生的基础真理方面始终比“k out of n”AV检测形式的所有未加权规则更准确。此外，我们评估了部分地真值可用于模型构建的场景。我们在部分标签信息上训练逻辑回归预测器。我们的结果表明，只需100个随机选择的训练实例，就足以达到80%的真阳性率和0.1%的假阳性率。相比之下，在假阳性率相同的情况下，最佳的非加权阈值规则只能提供60%的真阳性率。
{"title":"Better Malware Ground Truth: Techniques for Weighting Anti-Virus Vendor Labels","authors":"Alex Kantchelian, Michael Carl Tschantz, Sadia Afroz, Brad Miller, Vaishaal Shankar, Rekha Bachwani, A. Joseph, J. D. Tygar","doi":"10.1145/2808769.2808780","DOIUrl":"https://doi.org/10.1145/2808769.2808780","url":null,"abstract":"We examine the problem of aggregating the results of multiple anti-virus (AV) vendors' detectors into a single authoritative ground-truth label for every binary. To do so, we adapt a well-known generative Bayesian model that postulates the existence of a hidden ground truth upon which the AV labels depend. We use training based on Expectation Maximization for this fully unsupervised technique. We evaluate our method using 279,327 distinct binaries from VirusTotal, each of which appeared for the first time between January 2012 and June 2014. Our evaluation shows that our statistical model is consistently more accurate at predicting the future-derived ground truth than all unweighted rules of the form \"k out of n\" AV detections. In addition, we evaluate the scenario where partial ground truth is available for model building. We train a logistic regression predictor on the partial label information. Our results show that as few as a 100 randomly selected training instances with ground truth are enough to achieve 80% true positive rate for 0.1% false positive rate. In comparison, the best unweighted threshold rule provides only 60% true positive rate at the same false positive rate.","PeriodicalId":426614,"journal":{"name":"Proceedings of the 8th ACM Workshop on Artificial Intelligence and Security","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122677774","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
D. Fifield, A. Geana, Luis MartinGarcia, M. Morbitzer, J. D. Tygar
Differences in the implementation of common networking protocols make it possible to identify the operating system of a remote host by the characteristics of its TCP and IP packets, even in the absence of application-layer information. This technique, "OS fingerprinting," is relevant to network security because of its relationship to network inventory, vulnerability scanning, and tailoring of exploits. Various techniques of fingerprinting over IPv4 have been in use for over a decade; however IPv6 has had comparatively scant attention in both research and in practical tools. In this paper we describe an IPv6-based OS fingerprinting engine that is based on a linear classifier. It introduces innovative classification features and network probes that take advantage of the specifics of IPv6, while also making use of existing proven techniques. The engine is deployed in Nmap, a widely used network security scanner. This engine provides good performance at a fraction of the maintenance costs of classical signature-based systems. We describe our work in progress to enhance the deployed system: new network probes that help to further distinguish operating systems, and imputation of incomplete feature vectors.
{"title":"Remote Operating System Classification over IPv6","authors":"D. Fifield, A. Geana, Luis MartinGarcia, M. Morbitzer, J. D. Tygar","doi":"10.1145/2808769.2808777","DOIUrl":"https://doi.org/10.1145/2808769.2808777","url":null,"abstract":"Differences in the implementation of common networking protocols make it possible to identify the operating system of a remote host by the characteristics of its TCP and IP packets, even in the absence of application-layer information. This technique, \"OS fingerprinting,\" is relevant to network security because of its relationship to network inventory, vulnerability scanning, and tailoring of exploits. Various techniques of fingerprinting over IPv4 have been in use for over a decade; however IPv6 has had comparatively scant attention in both research and in practical tools. In this paper we describe an IPv6-based OS fingerprinting engine that is based on a linear classifier. It introduces innovative classification features and network probes that take advantage of the specifics of IPv6, while also making use of existing proven techniques. The engine is deployed in Nmap, a widely used network security scanner. This engine provides good performance at a fraction of the maintenance costs of classical signature-based systems. We describe our work in progress to enhance the deployed system: new network probes that help to further distinguish operating systems, and imputation of incomplete feature vectors.","PeriodicalId":426614,"journal":{"name":"Proceedings of the 8th ACM Workshop on Artificial Intelligence and Security","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129516903","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Differential privacy provides powerful guarantees that individuals incur minimal additional risk by including their personal data in a database. Most work in differential privacy has focused on differentially private algorithms that produce models, counts, and histograms. Nevertheless, even with a classification model produced by a differentially private algorithm, directly reporting the classifier's performance on a database has the potential for disclosure. Thus, differentially private computation of evaluation metrics for machine learning is an important research area. We find effective mechanisms for area under the receiver-operating characteristic (ROC) curve and average precision.
{"title":"Differential Privacy for Classifier Evaluation","authors":"Kendrick Boyd, Eric Lantz, David Page","doi":"10.1145/2808769.2808775","DOIUrl":"https://doi.org/10.1145/2808769.2808775","url":null,"abstract":"Differential privacy provides powerful guarantees that individuals incur minimal additional risk by including their personal data in a database. Most work in differential privacy has focused on differentially private algorithms that produce models, counts, and histograms. Nevertheless, even with a classification model produced by a differentially private algorithm, directly reporting the classifier's performance on a database has the potential for disclosure. Thus, differentially private computation of evaluation metrics for machine learning is an important research area. We find effective mechanisms for area under the receiver-operating characteristic (ROC) curve and average precision.","PeriodicalId":426614,"journal":{"name":"Proceedings of the 8th ACM Workshop on Artificial Intelligence and Security","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128133358","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Enterprise security is about protecting an enterprise's computing infrastructure and the enterprise's sensitive information stored and processed by the infrastructure. We secure the infrastructure and the information by combining three steps: (a) prevention, i.e., preventing security breaches to the extent possible, (b) detection, i.e., detecting breaches as soon as possible since prevention is not fool-proof, and (c) recovery, i.e., recovering from and responding to breaches after detection. Prior work, both in academia and in industry, has focused on prevention and detection, whereas recovery is a relatively unexplored area. Machine learning as a discipline has had a significant impact over the state of the art in enterprise security in the last few years, especially in the prevention and detection steps. However, widespread adoption remains a challenge for several reasons. In this talk, we describe current uses of machine learning in the prevention and detection steps, and highlight a few key challenges. We then discuss future opportunities for machine learning to improve the state of the art in recovery.
{"title":"Machine Learning for Enterprise Security","authors":"P. Manadhata","doi":"10.1145/2808769.2808782","DOIUrl":"https://doi.org/10.1145/2808769.2808782","url":null,"abstract":"Enterprise security is about protecting an enterprise's computing infrastructure and the enterprise's sensitive information stored and processed by the infrastructure. We secure the infrastructure and the information by combining three steps: (a) prevention, i.e., preventing security breaches to the extent possible, (b) detection, i.e., detecting breaches as soon as possible since prevention is not fool-proof, and (c) recovery, i.e., recovering from and responding to breaches after detection. Prior work, both in academia and in industry, has focused on prevention and detection, whereas recovery is a relatively unexplored area. Machine learning as a discipline has had a significant impact over the state of the art in enterprise security in the last few years, especially in the prevention and detection steps. However, widespread adoption remains a challenge for several reasons. In this talk, we describe current uses of machine learning in the prevention and detection steps, and highlight a few key challenges. We then discuss future opportunities for machine learning to improve the state of the art in recovery.","PeriodicalId":426614,"journal":{"name":"Proceedings of the 8th ACM Workshop on Artificial Intelligence and Security","volume":"58 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115619452","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
M. D. Cock, Rafael Dowsley, Anderson C. A. Nascimento, S. Newman
This work proposes a protocol for performing linear regression over a dataset that is distributed over multiple parties. The parties will jointly compute a linear regression model without actually sharing their own private datasets. We provide security definitions, a protocol, and security proofs. Our solution is information-theoretically secure and is based on the assumption that a Trusted Initializer pre-distributes random, correlated data to the parties during a setup phase. The actual computation happens later on, during an online phase, and does not involve the trusted initializer. Our online protocol is orders of magnitude faster than previous solutions. In the case where a trusted initializer is not available, we propose a computationally secure two-party protocol based on additive homomorphic encryption that substitutes the trusted initializer. In this case, the online phase remains the same and the offline phase is computationally heavy. However, because the computations in the offline phase happen over random data, the overall problem is embarrassingly parallelizable, making it faster than existing solutions for processors with an appropriate number of cores.
{"title":"Fast, Privacy Preserving Linear Regression over Distributed Datasets based on Pre-Distributed Data","authors":"M. D. Cock, Rafael Dowsley, Anderson C. A. Nascimento, S. Newman","doi":"10.1145/2808769.2808774","DOIUrl":"https://doi.org/10.1145/2808769.2808774","url":null,"abstract":"This work proposes a protocol for performing linear regression over a dataset that is distributed over multiple parties. The parties will jointly compute a linear regression model without actually sharing their own private datasets. We provide security definitions, a protocol, and security proofs. Our solution is information-theoretically secure and is based on the assumption that a Trusted Initializer pre-distributes random, correlated data to the parties during a setup phase. The actual computation happens later on, during an online phase, and does not involve the trusted initializer. Our online protocol is orders of magnitude faster than previous solutions. In the case where a trusted initializer is not available, we propose a computationally secure two-party protocol based on additive homomorphic encryption that substitutes the trusted initializer. In this case, the online phase remains the same and the offline phase is computationally heavy. However, because the computations in the offline phase happen over random data, the overall problem is embarrassingly parallelizable, making it faster than existing solutions for processors with an appropriate number of cores.","PeriodicalId":426614,"journal":{"name":"Proceedings of the 8th ACM Workshop on Artificial Intelligence and Security","volume":"85 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126230491","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Methods of compression-based text classification have proven their usefulness for various applications. However, in some classification problems, such as spam filtering, a classifier confronts one or many adversaries willing to induce errors in the classifier's judgment on certain kinds of input. In this paper, we consider the problem of finding thrifty strategies for character-based text modification that allow an adversary to revert classifier's verdict on a given family of input texts. We propose three statistical statements of the problem that can be used by an attacker to obtain transformation models which are optimal in some sense. Evaluating these three techniques on a realistic spam corpus, we find that an adversary can transform a spam message (detectable as such by an entropy-based text classifier) into a legitimate one by generating and appending, in some cases, as few additional characters as 11% of the original length of the message.
{"title":"Automated Attacks on Compression-Based Classifiers","authors":"Igor Burago, Daniel Lowd","doi":"10.1145/2808769.2808778","DOIUrl":"https://doi.org/10.1145/2808769.2808778","url":null,"abstract":"Methods of compression-based text classification have proven their usefulness for various applications. However, in some classification problems, such as spam filtering, a classifier confronts one or many adversaries willing to induce errors in the classifier's judgment on certain kinds of input. In this paper, we consider the problem of finding thrifty strategies for character-based text modification that allow an adversary to revert classifier's verdict on a given family of input texts. We propose three statistical statements of the problem that can be used by an attacker to obtain transformation models which are optimal in some sense. Evaluating these three techniques on a realistic spam corpus, we find that an adversary can transform a spam message (detectable as such by an entropy-based text classifier) into a legitimate one by generating and appending, in some cases, as few additional characters as 11% of the original length of the message.","PeriodicalId":426614,"journal":{"name":"Proceedings of the 8th ACM Workshop on Artificial Intelligence and Security","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121853751","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In the last several years, differential privacy has become the leading framework for private data analysis. It provides bounds on the amount that a randomized function can change as the result of a modification to one record of a database. This requirement can be satisfied by using the exponential mechanism to perform a weighted choice among the possible alternatives, with better options receiving higher weights. However, in some situations the number of possible outcomes is too large to compute all weights efficiently. We present the subsampled exponential mechanism, which scores only a sample of the outcomes. We show that it still preserves differential privacy, and fulfills a similar accuracy bound. Using a clustering application, we show that the subsampled exponential mechanism outperforms a previously published private algorithm and is comparable to the full exponential mechanism but more scalable.
{"title":"Subsampled Exponential Mechanism: Differential Privacy in Large Output Spaces","authors":"Eric Lantz, Kendrick Boyd, David Page","doi":"10.1145/2808769.2808776","DOIUrl":"https://doi.org/10.1145/2808769.2808776","url":null,"abstract":"In the last several years, differential privacy has become the leading framework for private data analysis. It provides bounds on the amount that a randomized function can change as the result of a modification to one record of a database. This requirement can be satisfied by using the exponential mechanism to perform a weighted choice among the possible alternatives, with better options receiving higher weights. However, in some situations the number of possible outcomes is too large to compute all weights efficiently. We present the subsampled exponential mechanism, which scores only a sample of the outcomes. We show that it still preserves differential privacy, and fulfills a similar accuracy bound. Using a clustering application, we show that the subsampled exponential mechanism outperforms a previously published private algorithm and is comparable to the full exponential mechanism but more scalable.","PeriodicalId":426614,"journal":{"name":"Proceedings of the 8th ACM Workshop on Artificial Intelligence and Security","volume":"303 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134454446","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
As antivirus and network intrusion detection systems have increasingly proven insufficient to detect advanced threats, large security operations centers have moved to deploy endpoint-based sensors that provide deeper visibility into low-level events across their enterprises. Unfortunately, for many organizations in government and industry, the installation, maintenance, and resource requirements of these newer solutions pose barriers to adoption and are perceived as risks to organizations' missions. To mitigate this problem we investigated the utility of agentless detection of malicious endpoint behavior, using only the standard built-in Windows audit logging facility as our signal. We found that Windows audit logs, while emitting manageable sized data streams on the endpoints, provide enough information to allow robust detection of malicious behavior. Audit logs provide an effective, low-cost alternative to deploying additional expensive agent-based breach detection systems in many government and industrial settings, and can be used to detect, in our tests, 83% percent of malware samples with a 0.1% false positive rate. They can also supplement already existing host signature-based antivirus solutions, like Kaspersky, Symantec, and McAfee, detecting, in our testing environment, 78% of malware missed by those antivirus systems.
{"title":"Malicious Behavior Detection using Windows Audit Logs","authors":"Konstantin Berlin, David Slater, Joshua Saxe","doi":"10.1145/2808769.2808773","DOIUrl":"https://doi.org/10.1145/2808769.2808773","url":null,"abstract":"As antivirus and network intrusion detection systems have increasingly proven insufficient to detect advanced threats, large security operations centers have moved to deploy endpoint-based sensors that provide deeper visibility into low-level events across their enterprises. Unfortunately, for many organizations in government and industry, the installation, maintenance, and resource requirements of these newer solutions pose barriers to adoption and are perceived as risks to organizations' missions. To mitigate this problem we investigated the utility of agentless detection of malicious endpoint behavior, using only the standard built-in Windows audit logging facility as our signal. We found that Windows audit logs, while emitting manageable sized data streams on the endpoints, provide enough information to allow robust detection of malicious behavior. Audit logs provide an effective, low-cost alternative to deploying additional expensive agent-based breach detection systems in many government and industrial settings, and can be used to detect, in our tests, 83% percent of malware samples with a 0.1% false positive rate. They can also supplement already existing host signature-based antivirus solutions, like Kaspersky, Symantec, and McAfee, detecting, in our testing environment, 78% of malware missed by those antivirus systems.","PeriodicalId":426614,"journal":{"name":"Proceedings of the 8th ACM Workshop on Artificial Intelligence and Security","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123599722","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：