We propose a system called NFVlambda where we will take a network wide holistic approach to building networks function utilizing a purely software approach. Each NF instead of being one big monolithic piece of software that is hard to debug, manage and extend will be composed of a multitude of different components that are specialized and perform a relatively simple task. But such components will be composable reused among different NFs and this will make building new NFs extremely efficient.

We envision that NFVlambda as a system that can be used to build scalable, extensible and manageable NFs using composable building blocks based on the actor model using functional programming that incorporates novel design features like packet specification for efficient and type correct packet processing. So that building new NFs for different scenarios becomes a hassle free and semantically correct process and hence eliminates the current difficulties faced by programmers.

NFVlambda is currently a project done in collaboration with Huawei and will be released as an open-source project for the community to build upon.


In our senior year project we have researched, designed and implemented Grizzly which is an automated malware detection framework based up virtual machine introspection. VMI gives us the ability to gather accurate information from the outside the system due to which it makes uses of the advantages of both HIDS and NIDS while avoiding their disadvantages. Grizzly provides automated signature based malware detection
based upon VMI where the signature can be defined in a high level language and hence the complexity is reduced as it is delegated to the framework. We believe that such frameworks can be deployed as as service in public clouds for malware detection using VMI.


SmarTor is a Tor implementation which delivers significant improvements over the current implementation. One of the main benefits of SmarTor is that it does not require any changes in the core infrastructure and only changes on the end client side are required. SmarTor leverages the user history and based on user profile optimally selects relays for circuit construction. Another aspect of SmarTor is that it reduces the entropy of the network by reducing the physical distance between the selected relays. Third, SmarTor takes a user defined variable between 0 and 1 which is a trade-off between latency and anonymity. The algorithm does all of this while taking precautionary measures to avoid traffic correlation attacks.