Who checks whether the system has a door that should not be open? Starting from this question, look at how a role connects needs, execution, and feedback.
Understanding today's problem
Understand what information the system handles, who has access, and where things might go wrong. Identify risks within an authorized scope, and confirm the impact with colleagues on the business side.
Moving forward with the team
Turn risks into improvements the team can carry out: access control, protection of sensitive information, log review, and recovery procedures. Work with engineering to confirm whether the approach is effective.
Delivery and continuous improvement
Review the fixes and record the risks and decisions. When an incident happens, collaborate by following the process, and gather the lessons so the same problem is less likely to happen again.
Role names help us see how work is divided; everyday collaboration is what truly connects those divisions.
A day in the role
The work, step by step
01Identify risks
Within an authorized scope, learn what information the system handles and who has access.
02Drive improvements
Turn risks into actionable improvements such as access control, protection of sensitive information, and log review.
03Incident response
When an incident happens, collaborate, record, and recover by following the process.
04Retrospective
Review the fixes so the same problem is less likely to happen again.
Across the eras
Where this work appears in the timeline
- Connection Era
1969–1993
Incident responders, with the starting point being the Morris worm.
- Web
1994–2000
Engineers at antivirus and firewall vendors, and members of the early hacker community.
- Community & Search Era
2001–2008
Web security engineers and penetration testers, doing vulnerability scanning, MLPS assessments, and incident response.
- Mobile Era
2009–2015
Mobile application security and business security (anti-cheating, anti-fraud), with enterprise SRCs collaborating with white-hat hackers.
- Cloud & Intelligence Era
2016–2022
Expanding from offense and defense into compliance. New data security and privacy compliance roles appeared (the DPO under GDPR, the person in charge of personal information protection domestically), working on zero trust and supply chain security.
- Large Model Era
2023–present
AI security (prompt injection, jailbreaks, the model supply chain), content security and filing compliance, and labeling of generated content.
Occupation records
Recognition, mappings and market data
U.S. pay and outlook use the Information Security Analysts occupation. The global WEF indicator uses the title Security Management Specialists.
Occupation names and classifications
China · Official occupation listing · A
Information Security Tester
Occupation listing · 2020.
人力资源社会保障部办公厅等（中国政府网）China · Official trade listing · A
Penetration Tester / Compliance Tester
Trade listing · 2021.
人力资源社会保障部办公厅等（中国政府网附件）China · Official occupation listing · A
Cybersecurity Classified Protection Assessor
Occupation listing · 2024.
人力资源社会保障部（中国政府网转载）China · Occupation classification / mapping · A
Data Security Engineering Technician
Classification code · 2-02-38-12.
中国就业网（人社部）
Pay, hiring and outlook indicators
United States · Proxy occupation · A
Information Security Analysts
Median annual pay： US$129,180
2025 median (P50) · USD/year
O*NET OnLine（美国劳工部资助） · 15-1212.00 Information Security Analysts
United States · Proxy occupation · A
Information Security Analysts
Projected growth / sample change： Much faster than average (7% or higher)
2025–2035 projected growth band
O*NET OnLine（美国劳工部资助） · 15-1212.00 Information Security Analysts
Global · Proxy occupation · B
Security Management Specialists
List position / share of job postings： #5
WEF Future of Jobs Report · 2025–2030 fastest net-growing roles list
Report list position; no in-chart percentage is shown.
World Economic Forum · The Future of Jobs Report 2025 (PDF)
Role record
History notes
These notes preserve the occupation names used by their sources. Era groupings and adjacent occupation names provide historical context; they do not mean those occupations are identical to this site’s role.
信息安全测试员 · Information Security Tester
The 2020 new-occupations release lists “信息安全测试员” (Information Security Tester). It is an occupation-name milestone in security, not a synonym for every security engineering role on this site.
数据安全工程技术人员 · Data Security Engineering Technician
The 2022 public notice lists “数据安全工程技术人员” (Data Security Engineering Technician), later included in the 2022 occupational classification dictionary. This is an official occupation-name milestone, not evidence that all security engineers began in that year.
网络安全等级保护测评师 · Cybersecurity Classified Protection Assessor
The 2024 new-occupations list from three government departments includes “网络安全等级保护测评师” (Cybersecurity Classified Protection Assessor). It is an official name for a specific security occupation, not a synonym for all security roles.
Related evidence · L1 · 30 papers
Contemporary research
These papers directly discuss responsibilities, skills, or tools relevant to the role. Topical relevance alone does not establish role history or replace employment and hiring data.
Towards Evaluating the Robustness of Neural Networks
Evaluating neural-network robustness against adversarial examples.
Abstract“vulnerable to adversarial examples”
Membership Inference Attacks against Machine Learning Models
Membership-inference attacks and privacy auditing of training data.
Title“Membership Inference Attacks”
The Limitations of Deep Learning in Adversarial Settings
Threat analysis of adversarial examples against deep-learning systems.
Title“adversarial settings”
Practical Black-Box Attacks against Machine Learning
Black-box attacks against machine-learning models.
Title“Practical Black-Box Attacks”
Extracting Training Data from Large Language Models
Training-data extraction risks for large language models.
Title“Extracting Training Data”
Fuzzy Extractors: How to Generate Strong Keys from Biometrics and Other Noisy Data
Biometric key generation and secure verification.
Title“Strong Keys from Biometrics”
Distillation as a Defense to Adversarial Perturbations against Deep Neural Networks
Defense mechanisms against adversarial perturbations.
Title“Defense to Adversarial Perturbations”
How To Backdoor Federated Learning
Backdoor threats in federated learning.
Title“Backdoor Federated Learning”
Spectre Attacks: Exploiting Speculative Execution
Spectre side-channel attacks and information disclosure.
Title“Spectre Attacks”
BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain
Model-supply-chain vulnerabilities and backdoor risks.
Title“Machine Learning Model Supply Chain”
Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoning
Data-poisoning and backdoor threats to deep-learning systems.
Title“Data Poisoning”
Evasion Attacks against Machine Learning at Test Time
Test-time evasion attacks and security evaluation.
Title“Evasion Attacks against Machine Learning”
RAPPOR: Randomized Aggregatable Privacy-Preserving Ordinal Response
Differential-privacy techniques for collecting client statistics.
Title“Privacy-Preserving Ordinal Response”
The Queen's Guard: A Secure Enforcement of Fine-grained Access Control In Distributed Data Analytics Platforms
Fine-grained access control for distributed data-analytics platforms.
Title“Fine-grained Access Control”
Stealing Machine Learning Models via Prediction APIs
Model-theft risks for public prediction APIs.
Title“Stealing Machine Learning Models via Prediction APIs”
Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection
Indirect prompt-injection risks in LLM-integrated applications.
Title“Indirect Prompt Injection”
Transferability in Machine Learning: from Phenomena to Black-Box Attacks using Adversarial Samples
Adversarial-sample transferability and black-box attacks.
Title“Black-Box Attacks using Adversarial Samples”
Celeb-DF: A Large-scale Challenging Dataset for DeepFake Forensics
Data and methods for deepfake forensics and detection.
Title“DeepFake Forensics”
Exploiting Unintended Feature Leakage in Collaborative Learning
Unintended information leakage in collaborative learning.
Title“Feature Leakage in Collaborative Learning”
Towards the Development of Realistic Botnet Dataset in the Internet of Things for Network Forensic Analytics: Bot-IoT Dataset
IoT intrusion detection and network forensics.
Title“Network Forensic Analytics”
Differentially Private Federated Learning: A Client Level Perspective
Client-level differential privacy for federated learning.
Title“Differentially Private Federated Learning”
Local Model Poisoning Attacks to Byzantine-Robust Federated Learning
Local model-poisoning attacks in federated learning.
Title“Local Model Poisoning Attacks”
Machine Unlearning
Data deletion and privacy-leakage risks.
Title“Machine Unlearning”
Renyi Differential Privacy
Analysis of differential-privacy guarantees.
Title“Renyi Differential Privacy”
Deep Models Under the GAN: Information Leakage from Collaborative Deep Learning
Information leakage in collaborative deep learning.
Title“Information Leakage from Collaborative Deep Learning”
A Survey on the Security of Blockchain Systems
Blockchain security threats and mitigation approaches.
Title“Security of Blockchain Systems”
N-BaIoT: Network-based Detection of IoT Botnet Attacks Using Deep Autoencoders
Anomaly detection for IoT botnet attacks.
Title“Detection of IoT Botnet Attacks”
A Survey on Homomorphic Encryption Schemes: Theory and Implementation
Homomorphic encryption for protecting sensitive cloud data.
Title“Homomorphic Encryption Schemes”
Fine-Pruning: Defending Against Backdooring Attacks on Deep Neural Networks
Defenses against backdoor attacks on neural networks.
Title“Defending Against Backdooring Attacks”
The Random Oracle Methodology, Revisited
Security analysis of cryptographic schemes.
Title“Random Oracle Methodology”