Computer Science and Information Technology
Course: CS818 — Doctoral Program in Computer Science (Cybersecurity & Information Assurance) Focus Area: Foundational CS Theory, Systems, Security, AI/ML, and Cloud Computing
Overview
This course provides a doctoral-level survey of core computer science and information technology domains — from the theoretical foundations of algorithms and computability, through operating systems, networking, databases, software engineering, computer graphics, artificial intelligence, and cloud security. All topics are examined through the lens of applied research in cybersecurity.
Computability and Algorithm Complexity
Functions and Computation
Computer science at its core asks: what can and cannot be computed?
- Computable functions can be expressed as a well-defined, step-by-step algorithm
- Noncomputable functions — no algorithm exists to determine outputs for all inputs; the halting problem is the canonical example
Turing Machines
Turing machines define the theoretical boundary of computation:
- A Turing machine reads symbols on a tape, transitions between states, and writes new symbols
- Any function computable by a modern computer can be computed by a Turing machine
- The halting problem — predicting whether any given program will terminate — is undecidable and cannot be solved by any algorithmic system
The Bare Bones Language
A minimal imperative programming language that can express all Turing-computable functions using only:
- Variables (non-negative integers as bit patterns)
- Three assignment statements
- One loop control structure (while)
This illustrates that computational power does not require complexity — minimalism can be Turing-complete.
Algorithm Complexity
Complexity measures how resource requirements (time, space) scale with input size.
| Notation | Meaning |
|---|---|
| Big-O (O) | Upper bound — worst-case growth rate |
| Big-Theta (Θ) | Tight bound — exact growth rate |
Time complexity counts execution steps, not lines of code.
| Algorithm | Complexity | Example |
|---|---|---|
| Binary Search | O(log n) | Finding a value in a sorted list |
| Merge Sort | O(n log n) | Efficient sorting |
| Bubble Sort | O(n²) | Naive sorting |
| Brute-force TSP | O(n!) | Traveling Salesperson Problem |
P vs. NP
| Class | Description |
|---|---|
| P (Polynomial) | Tractable — solvable in polynomial time |
| NP (Nondeterministic Polynomial) | Verifiable in polynomial time; may not be solvable quickly |
| NP-Complete | Hardest problems in NP; if one is solved in P, all NP problems are |
| Intractable | Cannot be solved in reasonable time for large inputs |
Open question: Does P = NP? The most important unsolved problem in theoretical computer science.
Heuristic approaches — for intractable problems, approximation algorithms find "good enough" solutions (e.g., greedy algorithms for TSP).
Advanced Persistent Threats as a High-Complexity Problem
APTs exemplify computational complexity in cybersecurity: - Multiple attack stages with branching decision paths - Customized exploits targeting specific vulnerabilities - Prolonged undetected presence (months to years) - Countermeasures: SIEM platforms, behavioral analytics, AI-based anomaly detection
Operating Systems
History and Evolution
| Era | Technology | Key Development |
|---|---|---|
| 1940s–1950s | Mainframes | Manual program setup; no OS |
| 1960s | Batch processing | Job queues (FIFO); job control language (JCL) |
| 1970s | Time-sharing | Multiple users; interactive processing via terminals |
| 1980s–1990s | Personal computing | Single-user multitasking; GUI interfaces |
| 2000s–present | Cloud & mobile | Multiprocessor, distributed, networked OS |
Core OS Functions
- Process management — Scheduling, dispatching, preventing interference between processes
- Memory management — Allocation, virtual memory, paging
- File management — Storage organization, access control, persistence
- Device drivers — Hardware abstraction layer
- User interface — CLI and GUI interaction
- Security — Authentication, privilege levels, access control
Multitasking and Time-Sharing
Multiprogramming creates the illusion of concurrent execution: - Multitasking (single user) — OS switches rapidly between processes - Time-sharing (multi-user) — CPU time is divided across users to ensure responsiveness
OS Comparison
| OS | Strengths | Weaknesses |
|---|---|---|
| Windows | Familiar UI; broad software compatibility; enterprise support | Licensing cost; more vulnerable to malware; resource-heavy |
| macOS | Seamless Apple integration; strong security; optimized hardware | Expensive hardware; limited customization; restricted compatibility |
| Linux | Free, open-source; highly secure; highly customizable | Steeper learning curve; less centralized support; some compatibility gaps |
Server OS Effectiveness
Windows Server: Integrates with Active Directory; familiar to Windows admins; strong enterprise support Linux Server: Preferred for web servers, databases, and cloud infrastructure due to low cost, stability, and security
OS Security Vulnerabilities
Malware and Ransomware - Viruses, worms, trojans, spyware encrypt or destroy files - Ransomware locks access and demands payment — disrupts operations and causes reputational damage
Zero-Day Vulnerabilities - Flaws unknown to the vendor — no patch exists at time of discovery - High-risk targets: Windows 10/11, Android (11–13), Windows Server 2012, Debian Linux
Most Vulnerable OS Versions (CVSS 7–10): Windows 10, Windows 11, Android 11–13, Windows Server 2012, Debian Linux 10/11, Fedora 37, HarmonyOS 2
Mitigation Strategies: - Enforce least-privilege access - Multi-factor authentication (MFA) - Patch management cadence - Endpoint detection and response (EDR) tools - Built-in CPU security features (hardware-enforced isolation)
Communication Protocols
Web Protocols and Their Security Roles
| Protocol | Purpose | Security Mechanism |
|---|---|---|
| HTTP | Data exchange between browser and server | None — plaintext |
| HTTPS | Secure HTTP | TLS/SSL encryption; Certificate Authority validation |
| DNS | Translates domain names to IP addresses | None inherently |
| DNSSEC | Authenticates DNS responses | Digital signatures prevent spoofing/cache poisoning |
| FTP | File transfer | None — plaintext |
| SFTP | Secure file transfer | SSH encryption |
| SMTP + STARTTLS | Secure email transmission | TLS upgrade for email channels |
HTTPS in Depth
- Browser sends HTTPS request to server
- Server presents TLS certificate (issued by a Certificate Authority)
- Browser verifies certificate authenticity
- TLS handshake establishes encrypted session parameters
- All subsequent data is encrypted — protected against eavesdropping, tampering, and man-in-the-middle attacks
CDN Security Role
Content Delivery Networks (CDNs) distribute assets globally, improving performance and providing DDoS protection by absorbing volumetric attack traffic before it reaches origin servers.
Impact of Network Loss
The COVID-19 pandemic demonstrated network dependency across every sector:
- Business — Remote work, video conferencing, e-commerce, cloud services would have been impossible without connectivity
- Education — Online platforms (Zoom, Google Classroom) sustained learning continuity
- Healthcare — Telemedicine replaced in-person consultations; vaccination data shared in real-time
- Government — E-governance services maintained during lockdowns
- Mental health — Social connectivity via online platforms reduced isolation impact
Algorithms and App Development
Algorithm Design Lifecycle (App Development)
- Ideation & Research — Define the problem; market analysis; set measurable goals
- Planning — Feature list; technology stack; timeline and budget
- Design — Wireframes → UI/UX design → interactive prototype
- Development — UI layer → backend/API → database integration → testing (unit, integration, UAT)
- Deployment — Containerize with Docker; push image to artifact registry; deploy to target platform
- Operations — Monitor performance; collect user feedback; release patches
Cybersecurity Algorithm Complexity
APTs require multi-step detection algorithms with high branching complexity: - Threat vector identification (branching per attack type) - Anomaly scoring across behavioral baselines - Correlation across SIEM event streams - Automated escalation or remediation decisions
Modern solutions: ML-powered SIEM, behavioral analytics, automated threat intelligence feeds
Security Issues and Prevention
Principal Software Vulnerabilities
SQL Injection - Attacker inserts SQL code into user input fields to manipulate database queries - Can read, alter, or delete data; execute OS commands - Detection: Static code analysis; penetration testing; ML classifiers (Naïve Bayes, decision trees, deep neural networks) - Mitigation: Parameterized queries / prepared statements; input validation; ORM frameworks
Cross-Site Scripting (XSS) - Malicious scripts injected into web pages viewed by other users - Mitigation: Output encoding; Content Security Policy (CSP) headers; input sanitization
Insecure Data Storage - Sensitive data stored without encryption or proper access controls - Mitigation: Encrypt data at rest (AES-256); enforce access control; audit storage configurations
Security-First Development Principles
- Embed vulnerability assessment into the SDLC — not as an afterthought
- Use static analysis tools (SAST) and dynamic analysis tools (DAST) at every build stage
- Conduct regular penetration testing with both automated tools and manual consultants
- Apply least-privilege access at the application, database, and OS layers
Software Development Frameworks
Framework Comparison
| Framework | Approach | Best For | Key Strength |
|---|---|---|---|
| Agile | Iterative (sprints ~2 weeks) | Evolving requirements; customer-driven | Flexibility; fast delivery; continuous feedback |
| Waterfall | Sequential (phase-gated) | Stable, well-defined requirements | Predictability; documentation; regulatory compliance |
| Kanban | Continuous flow | Ongoing maintenance and support | Visual workflow management; limit WIP |
| Lean | Waste elimination | Efficiency-focused teams | Maximize value; minimize non-essential work |
Choosing the Right Framework
- Agile — Use when requirements will change; customer collaboration is possible; fast time-to-market is critical
- Waterfall — Use when requirements are fixed; documentation is mandatory (healthcare, government, defense)
- Cybersecurity projects often benefit from Agile for threat response tools, Waterfall for compliance-driven security audits
Open-Source Software
Open Source vs. Licensed Software
| Dimension | Open Source | Licensed/Proprietary |
|---|---|---|
| Cost | Free (hidden costs: customization, support) | License fee (includes support, updates) |
| Customization | Fully customizable (source code access) | Limited; may cost extra |
| Support | Community-driven; paid options available | Vendor-backed; reliable professional support |
| Security | Rapid community patching; vulnerabilities publicly visible | Closed source; vendor-controlled patch timing |
| Compliance | License terms require legal review (GPL, MIT, Apache) | Clear vendor licensing terms |
Open Source in Cybersecurity and AI
- Cybersecurity: Open-source tools (Snort, OSSEC, OpenVAS) enable rapid vulnerability identification through global community peer review
- AI/ML: Open frameworks (TensorFlow, PyTorch, Hugging Face) accelerate research and deployment — enabling organizations to build, share, and improve models without vendor lock-in
- Tension: Transparency that enables rapid security improvement also exposes vulnerabilities to malicious actors — balance between openness and security protocols is essential
Data Storage and Manipulation
Data Compression
Lossless Compression — All data preserved; exact reconstruction guaranteed - Examples: ZIP, PNG, GIF - Used for: Code, documents, archives
Lossy Compression — Some data discarded for higher compression ratio - Examples: JPEG, MP3, MPEG - Used for: Images, audio, video where perceptual accuracy is sufficient
Compression Algorithm Comparison
| Algorithm | Compression Ratio | Speed | Best For |
|---|---|---|---|
| Brotli | Excellent | Moderate | Web assets (HTML, CSS, JS) |
| 7-Zip | Highest | Slow | Long-term archiving |
| LZ4 | Moderate | Fastest | Real-time, in-memory compression |
| WinZip | Good | Fast | General purpose; balanced |
Size Reduction Formula: Size Reduction = 1 - (1 / Compression Ratio)
WAN Optimization via Reference Indexing
A novel approach for optimal data transfer across Wide Area Networks:
- Analyze source data at the byte level
- Compare chunks to a reference database — generate metadata pointers instead of transmitting raw data
- Compress metadata using lossless techniques
- Transmit compressed metadata (significantly smaller than original)
- At destination, decompress and reconstruct using the reference table
Key advantage: Particularly effective in bandwidth-constrained environments; more flexible than traditional deduplication
Dissertation Topics in Data
- WAN Optimization — Reference indexing for efficient data center-to-data center transfer
- Real-Time Data Manipulation and Cybersecurity — AI-driven detection of data tampering in live streams
- Blockchain for Secure Data Storage — Immutable, distributed ledger for data integrity assurance
Databases
Relational Databases (RDBMS)
- Data organized in tables with predefined schemas
- Relationships enforced through foreign keys
- Query language: SQL
- ACID properties: Atomicity, Consistency, Isolation, Durability
| Database | Strengths | Common Use Cases |
|---|---|---|
| MySQL | Open-source; fast; widely supported | Web apps (WordPress, e-commerce, YouTube) |
| PostgreSQL | Advanced SQL; extensible; ACID-compliant | Complex analytics, financial systems |
| Oracle | Enterprise-grade; high availability | Banking, ERP, government |
| MS SQL Server | Windows integration; BI tools | Enterprise applications, .NET stacks |
NoSQL Databases
Designed for unstructured data, horizontal scalability, and high availability in distributed environments.
| Type | Example | Best For |
|---|---|---|
| Document | MongoDB | JSON data, content management, catalogues |
| Key-Value | Redis, DynamoDB | Sessions, caching, real-time leaderboards |
| Column-Family | Cassandra | Time-series, IoT, write-heavy workloads |
| Graph | Neo4j | Social networks, fraud detection, knowledge graphs |
Data Abstraction Levels
| Level | Description |
|---|---|
| Physical | How data is stored on disk (files, indexes, pages) |
| Logical | How data is structured (tables, relationships, schema) |
| View | How data is presented to specific users (filtered, role-based) |
Abstracting Dissertation Data for Cybersecurity
For AI cybersecurity research, raw data should be abstracted into: - Performance metrics table: Accuracy, false positives/negatives, response time, resource utilization - Threat landscape summary: Attack types, frequency, severity (Critical/High/Medium/Low) - Human-AI interaction themes: Trust levels, expert feedback, adoption barriers
Presentation tools: Bar charts (model accuracy comparison), line graphs (response time trends), thematic tables (qualitative interview themes)
Computer Graphics
Core Areas
- 2D Graphics — Raster (pixel-based: PNG, JPEG) and Vector (math-based: SVG, PDF)
- 3D Graphics — Modeling, rendering, lighting, texture mapping
- Animation — Keyframing, motion capture, procedural animation
- Simulation — Physics engines, particle systems, fluid dynamics
Industry Impact
Film and Entertainment - CGI enables photorealistic environments — Avatar, Jurassic Park, Toy Story - Real-time 3D rendering powers gaming (Unity, Unreal Engine) - Independent filmmakers empowered by accessible digital creation tools
Healthcare - 3D medical imaging (MRI, CT, ultrasound) provides detailed anatomical visualization - Surgical simulation training reduces risk in high-stakes procedures - Digital twins of organs enable personalized treatment planning
Graphics Processing Units (GPUs)
GPUs were designed for rendering but have become the dominant compute platform for AI, science, and data:
| Application | Why GPU Excels |
|---|---|
| Deep Learning / AI | Massive parallelism for matrix operations and backpropagation |
| Scientific Computing | Fluid dynamics, molecular modeling, particle physics simulations |
| Data Mining | Parallel processing of large datasets |
| Cryptocurrency Mining | High-throughput hash computations |
CPU vs. GPU Architecture: - CPU: Few powerful cores (4–64); optimized for sequential, low-latency tasks - GPU: Thousands of smaller cores; optimized for massively parallel workloads
Impact on AI: Training that takes days on a CPU completes in hours on a GPU. NVIDIA's Tensor Core (Volta architecture) specifically accelerates deep learning workloads.
Artificial Intelligence and Machine Learning
AI vs. ML Distinction
| Dimension | Artificial Intelligence (AI) | Machine Learning (ML) |
|---|---|---|
| Scope | Broad — all intelligent machine behavior | Subset of AI — learning from data |
| Dependency | Can exist without ML (rule-based systems) | Cannot exist without AI |
| Goal | Emulate human intelligence broadly | Learn patterns; improve predictions over time |
| Approach | Rules, reasoning, symbolic logic, ML | Data-driven; statistical models |
| Examples | Expert systems, robotics, NLP, computer vision | Classification, regression, clustering, deep learning |
ML Techniques
| Type | Description | Example |
|---|---|---|
| Supervised Learning | Trains on labeled input-output pairs | Spam detection, fraud classification |
| Unsupervised Learning | Finds patterns in unlabeled data | Clustering anomalies in network traffic |
| Reinforcement Learning | Learns by reward/penalty feedback | Adaptive intrusion response systems |
| Deep Learning | Multi-layer neural networks for complex patterns | Malware detection, image recognition |
AI Strengths and Weaknesses
Strengths: - Handles diverse tasks across domains - Excels at decision-making with incomplete information - Versatile — combines NLP, vision, reasoning, and learning
Weaknesses: - Requires significant compute, data, and tuning investment - Most AI is "narrow" — struggles to generalize across unrelated domains - AI trained on biased data produces biased outputs — critical concern in security profiling
AI in Cybersecurity — Dissertation Focus
Identified Gap: Ethical frameworks for AI in cybersecurity are underdeveloped; AI bias leads to unfair security outcomes (over-targeting specific user groups or attack types); zero-day threats remain a persistent challenge.
Proposed Research Topics: 1. Mitigating Bias and Enhancing Ethical Considerations in AI-Driven Cybersecurity Systems 2. Explainable AI (XAI) in Cybersecurity for Enhanced Threat Detection and Response
Research Methodology: - Review literature on AI threat detection, focusing on zero-day attacks - Develop a GAN-based cybersecurity framework — one network generates attack strategies, another detects them - Test against real-world cybersecurity datasets; compare to existing AI-based intrusion detection systems
Expected Contributions: - New methodology for integrating explainability into AI threat detection - Framework for human-AI collaboration in SOC environments - Adaptive security architecture that evolves with emerging threats
Cloud Computing and Security
Cloud Computing Fundamentals
Cloud computing delivers on-demand compute, storage, and networking resources over the internet — adaptable, durable, and cost-efficient.
Service Models: | Model | Description | Examples | |---|---|---| | IaaS | Infrastructure — VMs, networking, storage | AWS EC2, Azure VMs | | PaaS | Platform — runtime, middleware, databases | AWS RDS, Google App Engine | | SaaS | Software — complete applications | Salesforce, Google Workspace |
Deployment Models: Public, Private, Hybrid, Multi-Cloud
Cloud Security Challenges
The shared responsibility model — cloud provider secures infrastructure; customer secures their data, applications, and access — creates potential gaps when responsibilities are unclear.
Top Cloud Security Issues:
1. Data Breaches - Unauthorized access to sensitive information (PII, financial data, IP) stored in distributed cloud infrastructure - Causes: Misconfigured storage buckets, weak IAM policies, unpatched vulnerabilities - Dissertation research angle: AI-powered detection and prevention methods for cloud data breaches - Methodology: Quantitative analysis of breach incident data + AI/ML detection models
2. Insecure APIs - APIs are the primary interface for cloud-to-cloud and application-to-cloud communication - Vulnerable APIs expose backend systems to unauthorized access, data exfiltration, and injection attacks - Mitigation: OAuth 2.0 authentication, API key validation, rate limiting, Apigee API gateway security policies, mutual TLS
Cloud Security Research Potential
Each issue can be developed into a dissertation:
- Data Breach Research: Develop AI-powered intrusion detection for cloud environments; evaluate detection accuracy, false positive rates, and response times
- API Security Research: Propose ontological framework for cataloging and mitigating cloud API vulnerabilities dynamically
Course Summary and Key Takeaways
-
Computability defines limits — Understanding what computers cannot do is as important as what they can; the halting problem remains unsolvable
-
Operating systems are the security foundation — Patch management, privilege enforcement, and multi-factor authentication must be built into OS strategy, not bolted on
-
Protocols enable and protect communication — HTTPS, DNSSEC, SFTP, and TLS are not optional for any production system handling sensitive data
-
Algorithm complexity drives architectural decisions — Choosing the right data structure and algorithm class (P vs. NP-hard) determines whether a system scales
-
Open source accelerates innovation but requires governance — Transparency is a double-edged sword; security policies must accompany open adoption
-
Databases are not one-size-fits-all — RDBMS for transactional integrity; NoSQL for scale and flexibility; choose based on data structure and access patterns
-
GPUs transformed AI — The shift from CPU to GPU computing made modern deep learning possible; hardware architecture is inseparable from AI capability
-
AI in cybersecurity requires ethics and explainability — Biased models produce unjust outcomes; XAI is not optional when AI drives security decisions
-
Cloud security is shared — Understanding the boundary between provider and customer responsibility is the first step in building a secure cloud architecture
-
Every topic connects to the dissertation — Each CS domain informs the research on AI-driven cybersecurity — from algorithms that detect APTs, to databases that store threat intelligence, to cloud platforms that host AI models
References
- Brookshear, J. G., & Brylow, D. (2018). Computer Science: An Overview (13th ed.). Pearson.
- Russell, S., & Norvig, P. (2021). Artificial Intelligence: A Modern Approach (4th ed.). Pearson.
- Silberschatz, A., & Galvin, P. (1998). Operating System Concepts. Wiley.
- Patterson, D., & Hennessy, J. (2016). Computer Organization and Design. Morgan Kaufmann.
- Stallings, W. (2003). Data and Computer Communications. Prentice Hall.
- Fuggetta, A. (2003). Open source software — An evaluation. Journal of Systems and Software, 66(1).
- Ali, M., Khan, S. U., et al. (2015). Security in cloud computing: Opportunities and challenges. Information Sciences, 305.
- Zhang, W., Li, Y., et al. (2022). Deep learning for SQL injection detection. Computers & Security.