Weekly Commit #016
Seven weeks ago I started a Master's at IST while keeping a part-time software engineering job. The first period just ended, so this week is a recap instead of the usual single story.
The two projects that shaped it
DIDA: a distributed trading app where order matching is replicated with Paxos, built in Java and gRPC. We started from a provided baseline and added harder guarantees step by step, up to changing the replica group at runtime with Vertical Paxos. My takeaway: Paxos is elegant on a slide. In code, most of the work is in the edge cases, like a message that arrives late, arrives twice, or reaches a node that is no longer the leader.
CSF, “Operation DeepFish”: a digital forensics investigation in three stages, run from a Kali Linux VM, each stage looking at the same fictional case from a different angle.
- Stage I, the artifacts: a phishing clone of the university login page. The server exposed its own .git folder, so we rebuilt the repo, including deleted files, and found eight files hidden inside images, video, audio and a spreadsheet. The hardest ones were in places you wouldn't check first: the difference between two audio channels, an animated image's frame timings, a JPEG's compression coefficients.
- Stage II, the endpoints: the suspect's seized laptop and workstation, memory and disk. The BitLocker key was still in RAM. He had securely wiped his files, but copies survived in the thumbnail cache, and his shell history told most of the story.
- Stage III, the network: packet captures from the public agency behind the AI system, with TLS decrypted from key logs. We traced how the files actually left the agency. The person who looked guilty from the start may have been framed.
My takeaways: every claim has to point to a piece of evidence, an account or an IP address is not a person, and anti-forensics is almost never complete.
What a week looked like
Most days followed the same pattern: classes in the morning, work in the afternoon, studying at night and on weekends. Some days broke it. There were days filled entirely with classes, and days where I worked from morning to evening, and either way the studying still happened at night. It wasn't always easy, but having a fixed rhythm is what made both fit.
Smaller things this period
- Presented the FlexLog paper (a shared log for stateful serverless computing) with two colleagues
- Communication Skills: three oral assessments, which were more useful than I expected
- Performance Evaluation and Dimensioning of Networks and Systems: learned to see systems as queues, where how long a request waits matters as much as how fast it's served. The course runs all semester, and the second period covers Kubernetes, server farms and more, which I'm looking forward to.
Period two starts now, with the same plan: one commit at a time. See you next week.
#WeeklyCommit #DistributedSystems #Cybersecurity #SoftwareEngineering #IST
Follow along here or on LinkedIn where I'll be sharing each new commit.