True/False – GK Questions

Service mesh (Istio, Linkerd) deploys sidecar proxies for mTLS, retries, metrics, tracing. Decouples infrastructure concerns from application code. Critical for cloud-native architectures.

Canary deployment minimizes risk by gradually expanding model exposure, monitoring metrics, and rolling back if issues arise. Critical for safe ML system updates in production.

LoRA adds low-rank matrices to weight updates, reducing trainable parameters by 100-1000x vs full fine-tuning. Enables efficient adaptation of LLMs on consumer hardware. Critical for resource-constrained AI.

Data localization (RBI payment data, DPDP Act provisions) ensures law enforcement access, reduces foreign surveillance risks. Balances sovereignty with global data flow needs. Critical for policy questions.

XAI methods (SHAP, LIME, attention visualization) provide interpretable explanations for AI decisions. Critical for regulated sectors (finance, healthcare) and user acceptance.

Serverless (Lambda, Cloud Functions) uses pay-per-execution pricing, auto-scaling from zero. Trade-offs: cold starts, execution limits, vendor lock-in. Critical for cost-optimized cloud architectures.

Deception deploys fake assets (servers, credentials) to lure attackers, enabling early detection and threat intelligence. Critical for proactive defense against advanced adversaries.

Feature stores (Feast, Tecton) manage feature engineering, versioning, and serving, preventing training-serving skew. Critical for reliable ML system performance in production.

Prompt injection exploits LLM instruction-following to bypass safety guards, extract training data, or execute unauthorized actions. Critical vulnerability for LLM applications requiring robust input sanitization.

Digital diplomacy: cyber norms, data governance, tech partnerships, capacity building. India actively engages in UN, GPAI, OECD on AI ethics, cybersecurity, digital trade. Critical for foreign policy questions.

Responsible data science: fairness audits for bias, privacy-preserving techniques, impact assessments. Critical for ethical AI deployment and regulatory compliance (DPDP Act).

IaC (Terraform, CloudFormation) defines infrastructure in declarative code, enabling version control, peer review, automated testing, and consistent deployments. Critical for cloud governance.

SOAR (Security Orchestration, Automation, and Response) automates playbooks: alert triage, enrichment, containment actions. Reduces analyst workload and response time. Critical for SOC efficiency.

Threat intelligence (commercial, open-source, ISACs) provides IOCs: malicious IPs, domains, hashes, TTPs. Enables proactive blocking and hunting. Critical for mature security operations.

Quantization (INT8, INT4) reduces model size 2-4x with minimal accuracy loss, enabling deployment on edge devices and reducing cloud inference costs. Critical for scalable LLM serving.

PEFT methods (LoRA, Adapter, Prefix-tuning) update small subset of parameters or add lightweight modules, reducing compute/memory vs full fine-tuning. Critical for resource-constrained LLM adaptation.

Fine-tuning updates pre-trained model weights on task-specific data, enabling adaptation to domains like healthcare, legal, or customer service with limited labeled examples. Critical for practical LLM deployment.

Internet is the global network of interconnected computers using TCP/IP protocols. WWW (World Wide Web) is a service that runs on the internet, using HTTP/HTTPS to access interlinked hypertext documents. Other internet services include email, FTP, VoIP. Web is a subset of internet.

CoWIN is built on a modular, scalable architecture that can be repurposed for other immunization and health programs. The government has indicated plans to adapt CoWIN for routine vaccination drives, demonstrating the value of building flexible digital public infrastructure for healthcare delivery.

No antivirus can guarantee 100% detection due to zero-day exploits, polymorphic malware, and advanced persistent threats. Effective security requires layered approach: updated AV, firewall, patches, user awareness, backups, and behavioral analysis. Regular updates improve detection but cannot eliminate all risks.