Now showing 1 - 2 of 2
- PublicationBehavioral Service Graphs: A Formal Data-Driven Approach for Prompt Investigation of Enterprise and Internet-wide InfectionsThe task of generating network-based evidence to support network forensic investigation is becoming increasingly prominent. Undoubtedly, such evidence is significantly imperative as it not only can be used to diagnose and respond to various network-related issues (i.e., performance bottlenecks, routing issues, etc.) but more importantly, can be leveraged to infer and further investigate network security intrusions and infections. In this context, this paper proposes a proactive approach that aims at generating accurate and actionable network-based evidence related to groups of compromised network machines (i.e., campaigns). The approach is envisioned to guide investigators to promptly pinpoint such malicious groups for possible immediate mitigation as well as empowering network and digital forensic specialists to further examine those machines using auxiliary collected data or extracted digital artifacts. On one hand, the promptness of the approach is successfully achieved by monitoring and correlating perceived probing activities, which are typically the very first signs of an infection or misdemeanors. On the other hand, the generated evidence is accurate as it is based on an anomaly inference that fuses data behavioral analytics in conjunction with formal graph theoretic concepts. We evaluate the proposed approach in two deployment scenarios, namely, as an enterprise edge engine and as a global capability in a security operations center model. The empirical evaluation that employs 10 GB of real botnet traffic and 80 GB of real darknet traffic indeed demonstrates the accuracy, effectiveness and simplicity of the generated network-based evidence.
282Scopus© Citations 6
- PublicationImproving Borderline Adulthood Facial Age Estimation through Ensemble LearningAchieving high performance for facial age estimation with subjects in the borderline between adulthood and non-adulthood has always been a challenge. Several studies have used different approaches from the age of a baby to an elder adult and different datasets have been employed to measure the mean absolute error (MAE) ranging between 1.47 to 8 years. The weakness of the algorithms specifically in the borderline has been a motivation for this paper. In our approach, we have developed an ensemble technique that improves the accuracy of underage estimation in conjunction with our deep learning model (DS13K) that has been fine-tuned on the Deep Expectation (DEX) model. We have achieved an accuracy of 68% for the age group 16 to 17 years old, which is 4 times better than the DEX accuracy for such age range. We also present an evaluation of existing cloud-based and offline facial age prediction services, such as Amazon Rekognition, Microsoft Azure Cognitive Services, How-Old.net and DEX.
200Scopus© Citations 10