Naučni radovi


Savremene visokotehnološke pretnje: o ranjivosti softverskih proizvoda i pretnjama


U ovom radu izvršena je analiza savremenih visokotehnoloških pretnji koje nastaju kao posledica sigurnosnih propusta u softveru i ranjivosti softverskih proizvoda. Ranjivosti u softverskim proizvodima često nastaju kao posledica primene metodologije brzog razvoja i predstavljaju pretnje koje napadači sa odgovarajućim znanjem i računarskim resursima mogu da iskoriste kako bi stekli neovlašćeni pristup računarskim sistemima i mrežama …


Autori

  • Miloš Jovanović
  • Nemanja Maček
  • Igor Franc
  • Dragan Mitić

Datum objave: 03/06/2016


An Approach to Robust Biometric Key Generation System Design


This paper presents a novel approach to the design of robust multimodal biometric cryptosystems. The design objectives behind the system are robustness, privacy of user’s biometric templates and stable cryptographic key generation. The framework presented in this paper employs two modalities and a look-up table …


Autori

  • Nemanja Maček
  • Borislav Đorđević
  • Jelena Gavrilović
  • Komlen Lalović

Datum objave: 01/12/2015


Performance Issues in Cloud Computing: KVM Hypervisor’s Cache Modes Evaluation


This paper examines the performance of bare-metal hypervisors within the context of Quality of Service evaluation for Cloud Computing. Special attention is paid to the Linux KVM hypervisors’ different cache modes. The main goal was to define analytical models of all caching modes provided by hypervisor according to the general service time equation …


Autori

  • Borislav Đorđević
  • Nemanja Maček
  • Valentina Timčenko

Datum objave: 01/07/2015


An Efficient Approach to Generating Cryptographic Keys from Face and Iris Biometrics Fused at the Feature Level


This paper present an efficient approach to secure cryptographic key generation from iris and face biometric traits. Features extracted from preprocessed face and iris images are fused at the feature level and the multimodal biometric template is constructed from the Gabor filter and Principal Component Analysis outputs …


Autori

  • Saad Abuguba
  • Milan Milosavljević
  • Nemanja Maček

Datum objave: 20/06/2015


Improving Intrusion Detection with Adaptive Support Vector Machines


The research topic that this paper is focused on is intrusion detection in critical network infrastructures, where discrimination of normal activity can be easily corrected, but no intrusions should remain undetected. The IDS presented in this paper is based on SVM that classify unknown data instances according both to the feature values and weight factors that represent importance of features towards the classification …


Autori

  • Nemanja Maček
  • Milan Milosavljević
  • Borislav Đorđević
  • Valentina Timčenko
  • Mladen Bojović

Datum objave: 01/12/2014


Reducing U2R and R2L Category False Negative Rates with Support Vector Machines


The KDD Cup ’99 is commonly used dataset for training and testing IDS machine learning algorithms. Some of the major downsides of the dataset are the distribution and the proportions of U2R and R2L instances, which represent the most dangerous attack types, as well as the existence of R2L attack instances identical to normal traffic. This enforces minor category detection complexity and causes problems while building a machine learning model capable of detecting these attacks with sufficiently low false negative rate …


Autori

  • Nemanja Maček
  • Milan Milosavljević

Datum objave: 01/02/2014


Kritička analiza KDD Cup’99 skupa podataka i metodologije istraživanja mašinskog učenja u IDS sistemima


U ovom radu data je kritička analiza skupa podataka KDD Cup ’99 i metodologije istraživanja mašinskog
učenja u oblasti detekcije upada zasnovanih na skupu. Iako se ovaj skup najčešće koristi za obuku, validaciju i testiranje algoritama za mašinsko učenje u IDS sistemima, podaci u skupu nisu verodostojna reprezentacija saobraćaja realne računarske mreže …


Autori

  • Nemanja Maček
  • Milan Milosavljević

Datum objave: 06/06/2013


Tehnike manipulacije slikom i detekcija copy-move napada


Pojavom računara i različitog softvera za obradu slika postalo je jednostavno falsifikovati ih. Kao
jedan od najuspešnijih primena metoda analize slike je detekcija modifikacije digitalne fotografije. Iako postoje različiti sistemi za detekciju manipulacije digitalnom fotografijom njihov uspeh je ograničen i još uvek je u razvoju. Ovaj rad je napisan u pokušaju da pomogne ovim naporima tako što će dati kratak pregled poznatih metoda kao i smernice za dalje istraživanje u ovoj oblasti.


Autori

  • Igor Franc
  • Miloš Stojmenović

Datum objave: 22/11/2012