TDDD17 Information Security, Second Course

Project Assignments

Project id	Project name	Assigned students	Supervisor
001	Local optima in evolutionary fuzzing	Group A: robek274, gusel275 Group B: augsv102, danmy683	Ulf Kargén
002	Android obfuscation and anti-analysis techniques	Group A: aleno523, marha057 Group B: hyagi891, adahy344	Ulf Kargén
003	Survey of mobile forensics	Group A: vikno500, tonli746 Group B: mayhe310, gabhu204	Ulf Kargén
004	Blockchain Security for IoT	Group A: marla070, maxmo027 Group B: frete700, fraoh005	Andrei Gurtov
005	Next-generation aviation cybersecurity	Group A: abdzu618, souja450 Group B: joath015, hanli523	Andrei Gurtov
006	Analyzing Industrial Devices	Group A: rebli574, emica716 Group B: rassa328, axeka694	Andrei Gurtov
007	Airwall Teams	Group A: julwa850 Group B: jones726, maxsk293	Andrei Gurtov
008	Cybersecurity study of a drone	Group A: antoo433, vikro145 Group B: chrek644, isach408	Andrei Gurtov
009	Features that make a design pattern "dark"	Group A: teoar379, gonve175 Group B: simhe569	Jenni Reuben
010	Curious case of privacy preserving machine learning solutions	Group A: fella149, joaos226 Group B:	Jenni Reuben
011	Analysis of Twitter Users' Perceptions of ChatGPT and Security	Group A: basal573, gusbo010 Group B: adian611, vikan595	Alireza Mohammadinodooshan
012	Quantifying the Role of Twitter Bots in Amplifying Tweet Interactions	Group A: ernla111, vilkn620 Group B: isagr354, hilfr291	Alireza Mohammadinodooshan

1. Local optima in evolutionary fuzzing

Fuzzing is a widely deployed random testing technique for automatically finding critical security flaws in software. Most fuzzers today follow the greybox paradigm, where mutated inputs that trigger new code coverage is saved in the so-called seed queue, to be re-fuzzed at the next pass over the queue. This way, fuzzers implement a (conceptually) simple evolutionary algorithm, which allows incremental discovery of new code paths and program behavior, and which drastically improves performance over simple black-box random testing. However, all evolutionary algorithms are prone to falling into local optima, which are then difficult to get out of. In the case of fuzzing, this could manifest itself as the fuzzing results (code coverage, number of bugs found) becoming strongly dependent on the particular code paths that the fuzzer happen to "stumble upon" during the first few minutes of fuzzing. This de-diversification of the search space leads to more unpredictable fuzzer performance (i.e., greater variance in between runs on the same software), and overall reduced likelihood of finding bugs.

The purpose of this project is to perform a preliminary empirical study using one or several state-of-the-art fuzzers, to understand to what degree this is a problem in practice. Concretely, a number of short fuzzing runs will first be performed. The discovered inputs of those runs will then be used as starting points (seeds) for a number of longer runs, and a statistical analysis will be performed to see if fuzzing runs with the same starting point exhibit a greater degree of similarity over time (coverage, bugs), compared to runs with different starting points. Computing resources for doing the experiments will be provided by us.

Prerequisites: Some familiarity with Linux systems (Bash scripting, etc.). Basic knowledge of fuzzing and software security (e.g., from TDDC90) is recommended, but not strictly necessary. Basic knowledge of statistics as well as Python programming skills are also recommended.

2. Android obfuscation and anti-analysis techniques

Obfuscation and anti-analysis techniques are common challenges faced by forensic analysts when reverse engineering malicious apps, or analyzing third-party apps for, e.g., security or privacy violations. The OWASP foundation has released a number of "crackmes" — hacking challenges — that exemplify several such anti-analysis techniques. The purpose of this project is to survey common anti-analysis techniques used in Android apps, and techniques and tools to circumvent them. In the report, you are expected to demonstrate select techniques and tools using the OWASP crackmes. Furthermore, you are also expected to perform at least one such demonstration during the final presentation.

https://github.com/OWASP/owasp-mstg/tree/master/Crackmes
https://mobile-security.gitbook.io/mobile-security-testing-guide/android-testing-guide/0x05j-testing-resiliency-against-reverse-engineering

Prerequisites: Good programming skills and familiarity with developing and debugging Android apps. Some knowledge of software security (e.g., from TDDC90) is recommended.

3. Survey of mobile forensics

Forensic analysis of mobile phones has become extremely important in, for example, criminal investigations and investigations of cyber attacks. The purpose of this project is to survey literature on mobile forensics (academic papers, internet resources, etc.) to answer the following questions: what are the overall processes, strategies and workflows commonly used when performing a mobile forensic investigation? What are the main challenges? What techniques and tools (hardware and software) are commonly used?

Prerequisites: Some familiarity with mobile app development probably helps, but isn't strictly necessary.

4. Blockchain Security for IoT

In this project, the goal is to explore blockchain usage scenarios in Internet of Things (IoT). A demo of some security-related scenario based on smart contracts should be setup. You can use several Raspberry PIs 3B/4/Zeros as test IoT devices. Public cloud for blockchain service could be setup, e.g. on Amazon AWS. Which Hyperledger (Fabric, Iroha, Indy, ...) is best as a blockchain platform? Simple starting scenarios could include e.g. adding or removing an IoT device to a SmartHome, or hosting a guest device. Can the system be applied as a registry for drone or vessel identification numbers? The report could include a literature review, prototype description and performance evaluation.

https://www.hyperledger.org
https://dl.acm.org/doi/10.1145/3479243.3487305
https://www.ledgerinsights.com/blockchain-drone-unmanned-aviation-skygrid/

Prerequisites: Knowledge of networking, distributed systems and basic Internet security.

5. Next-generation aviation cybersecurity

L-band Digital Aeronautical Communications System (LDACS) and Aeronautical Mobile Airport Communication System (AEROMACS), SatCom are three new systems being developed for future data communication (FCI) for civil aviation. Those are meant to replace current systems for data communication and navigation, such as ADS-B, CPDLC, ACARS, etc which are insecure. The goal of the project is to study the security mechanisms of LDACS, SatCom, AEROMACS, estimate their strength, overhead, capability to interoperate. How to provide smooth handovers between these technologies, is Mobile IP or Host Identity Protocol (HIP) useful here? This is mostly a literature study project.

https://datatracker.ietf.org/doc/draft-ietf-raw-ldacs/
https://www.eurocontrol.int/function/future-communications-infrastructure-and-multilink-long-term
https://corinna-schmitt.de/publications-conference.html
https://wimaxforum.org/Page/AeroMACS

Prerequisites: Knowledge of networking and basic Internet security and wireless protocols.

6. Analyzing Industrial Devices

Many Internet-of-things devices (IP cams, smart TVs, etc) and industrial equipment including so-called SCADA devices are being connected to the Internet. In many cases, those are not patched, use default passwords or are not supposed to be on the Internet at all. Those can (and were already) easily hacked and exploited by attackers. Shodan is a search engine that allows to look for such devices in a certain area (identified by IP address ranges). The instructor will provide a paid Shodan account to use.

The focus is on identifying interesting devices connected to the Internet in EU. Do some exploration and testing to include interesting device pictures and descriptions to the report. A description of trends compared to past years should be included as well. Can Shodan discover devices connected with 4G, Sigfox or other IoT type networks? How to hide your devices from discovery?

https://www.shodan.io/explore/category/industrial-control-systems
https://aaltodoc.aalto.fi/bitstream/handle/123456789/12918/master_Tiilikainen_Seppo_2014.pdf?sequence=1
https://ieeexplore.ieee.org/document/8372775

Prerequisites: Knowledge of networking and basic Internet security.

7. Airwall Teams

Airwall Teams is a security platform for micro-segmentation and remote access developed by Tempered Networks. According to the marketing material on their webpage: "Airwall Teams does the impossible; allows you to build truly private system-to-system networks—that span public, private, cloud, and mobile networks—with just a few clicks using an intuitive graphical interface. Traverse NAT, firewalls, and other obstacles using our standards-based Host Identity Protocol agents and ignite your team's productivity. Airwall Teams is a powerful subset of Enterprise and Industrial network solution, Airwall—and it's free."

The goal of this project is to study this system and setup a testbed to experiment with it, measure performance (throughput, delay, loss, etc).

https://www.tempered.io/products/airwall-teams/
https://newsdirect.com/news/tempered-networks-launches-airwall-teams-free-zero-trust-remote-access-and-private-network-solution-418782508 https://www.tempered.io/assets/pdfs/primer-hip-whitepaper-2.pdf

Prerequisites: Knowledge of networking and basic Internet security.

8. Cybersecurity study of a drone

The goal of this project is to study potential cybersecurity issues in a commercial drone, such as DJI Phantom 4. In the start, you should check existing literature on documented attacks e.g. on GPS or WiFi jamming. Then you can try to reproduce some attacks using Software Define Radio (SDR) in a safe controlled environment. Another related issue is broadcast remote ID for drones. You can explore which format and data the drones are currently broadcasting and if there are any potential privacy concerns. Also if it is compatible with current standards by ASTM and U-Space. A Drone and SDR testing hardware will be available to use.

http://kth.diva-portal.org/smash/get/diva2:1463784/FULLTEXT01.pdf
https://arxiv.org/pdf/2207.10795
https://datatracker.ietf.org/wg/drip/about/
https://www.astm.org/f3411-22a.html

Prerequisites: Knowledge of networking and basic Internet security.

9. Features that make a design pattern "dark"

GDPR mandates that any form of personal data collection by a service provider should be preceded with an informed consent given by the subject of whom the personal data concerns. Anti-privacy patterns or simply "dark patterns" are UI design elements in the web forms that make the users to take actions that they would have not otherwise consciously do in the absence of dark patterns. Many studies have been conducted to validate the idea that dark patterns in cookie banners are used to elicit consent. Consequently, research efforts to detect dark patterns using machine learning based approaches have risen since then. This project is about designing a study to uncover important training features to detect dark patterns in cookie banners. The focus of the project is the method to synthesize the important features that exists in the literature, and the classification of the identified features.

Prerequisites: Knowledge of HTTP and web development. Basic knowledge of machine learning.

10. Curious case of privacy preserving machine learning solutions

Data silos (i.e., data repositories that are inaccessible to all but a small group of people in an organization) undermine the generalization of machine learning models, for example in medical image analysis, recommendation systems, etc. Data sharing for the purpose of centralized training is not always possible due to various legal, business and technical reasons. Privacy Preserving Machine Learning (PPML) is a research discipline that enable two or more data owners to train a global model from private intermediate training results computed locally by each owner from their respective data samples. Research in PPML approaches shows two different research directions: 1) differential privacy-guaranteed aggregation of a global model and 2) homomorphic encryption-based computation of a global model. In this project the threat model, guarantees, trade-offs and performances of these two specific application (one from each category) are studied using the use case method.

Prerequisites: Basic knowledge of machine learning and encryption. Programming experience in Python is required.

11. Analysis of Twitter Users' Perceptions of ChatGPT and Security

This project is motivated by the growing importance of AI models in a variety of fields and the need to understand people's perspectives and opinions regarding the security aspects of these models. In this regard, this project seeks to investigate the perspectives and concerns of individuals regarding "security and ChatGPT", a sophisticated AI language model developed by OpenAI. The research will involve collecting tweets that mention ChatGPT and security and analyzing them manually and semi-manually to identify common themes and patterns in people's perspectives. Twitter data will be collected using relevant keywords, and the collected tweets will be manually and automatically analyzed using ready-to-use NLP (natural language processing) packages to identify themes and patterns in people's perspectives on ChatGPT and security. For instance, we will determine which security concepts and aspects people are most concerned about when using these tools.

Prerequisites: Good programming skills, basic statistics. Basic familiarity with sentiment analysis and topic modeling will be a merit, yet they are not mandatory as ready-made packages will be used.

12. Quantifying the Role of Twitter Bots in Amplifying Tweet Interactions

The use of social bots in social media is a growing concern for online security. An estimated 9 to 15 percent of Twitter users are bots, and the proliferation of low-quality content, such as fake news, is one of the greatest threats to social security. Through their actions, such as likes and shares, social media users play an important role in the dissemination of this content. Bots are a subset of these users designed to facilitate the dissemination of specific content. The purpose of this study is to determine the extent to which Twitter bots contribute to engagement with different types of tweets. For this purpose, we will collect recent tweets from a list of publishers with known biases and credibility and examine the interactions of a representative sample of users with those tweets and the roles Bots play here. There are websites, such as Media Bias Fact Check, that provide labels for publishers. The tweets can be retrieved using either the Twitter API or third-party applications such as Tweepy. The bot likelihood of users will also be extracted from the Botometer tool, using its free API.

Prerequisites: Good programming skills, basic statistics.

 

Page responsible: Ulf Kargén
Last updated: 2023-02-06