A Taxonomy of Attacks and a Survey of Defence Mechanisms for Semantic Social Engineering Attacks
The DOI matches its Crossref record; 149 of the 153 references it cites were found in a registry or an index; 9 of the 131 sources its claims cite were checked for support.
Abstract
Social engineering is used as an umbrella term for a broad spectrum of computer exploitations that employ a variety of attack vectors and strategies to psychologically manipulate a user. Semantic attacks are the specific type of social engineering attacks that bypass technical defences by actively manipulating object characteristics, such as platform or system applications, to deceive rather than directly attack the user. Commonly observed examples include obfuscated URLs, phishing emails, drive-by downloads, spoofed websites and scareware to name a few. This article presents a taxonomy of semantic attacks, as well as a survey of applicable defences. By contrasting the threat landscape and the associated mitigation techniques in a single comparative matrix, we identify the areas where further research can be particularly beneficial.
Overview
CoChat Summary
TL;DR
This survey proposes a taxonomy for semantic social engineering attacks and maps it to defence mechanisms in a comparative attack-vs-defence matrix. The paper argues that current defences are fragmented and that combining awareness, sandboxing, machine learning, and policy controls could better cover the semantic attack space.
Key findings
- The taxonomy classifies attacks across three control stages: orchestration, exploitation, and execution, using parameters such as target description, method of distribution, mode of automation, deception vector, interface manipulation, attack persistence, and execution steps.
- The paper states that its taxonomy is based on 30 different attacks observed in the wild and is intended to represent both existing and future attacks by focusing on shared characteristics rather than specific vectors.
- In the attack-vs-defence matrix, the authors say promiscuous targeting (TD2), software-based remote distribution (MD1-R), and automation (MA2) are common across most attacks, and that a defence against this orchestration pattern could apply to almost half of the listed semantic attack types.
- The paper reports several specific evaluation numbers from prior work it surveys, including 99% accuracy for SURF, 91% accuracy with 0.87% false positive margin for one URL filtering system, 92.52% accuracy for a Twitter phishing detector, and a false positive rate below 0.05% for a behavioral email model with 7000 emails and over in the user history.
Methods
- Literature survey and taxonomy design: the authors build a linear, single-layer taxonomy of semantic attacks using three stages of control and mutually exclusive classification criteria.
- Comparative mapping: they construct an attack-vs-defence matrix that maps surveyed defence mechanisms to the taxonomy’s categories.
- Use of examples: the paper illustrates the taxonomy with detailed attack examples and summarizes observed attacks in tables, including the 30-attack table and taxonomy classifications for specific attack families.
Limitations
- The paper explicitly says the taxonomy is not exhaustive.
- It states that there is no authoritative, up-to-date, publicly available dataset for semantic attacks, and that collecting such data is difficult because of the human factor and ethical constraints in deception-based research.
- The text makes evident that much of the survey is based on prior literature rather than a direct experimental evaluation of the taxonomy itself (evident from the paper’s survey-and-matrix structure).
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures








Verification
Claims the paper makes
8 Valid · 15 Not valid · 81 Not checked · 94 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“In [Jagatic et al. 2007], the phishing attack contains three separate steps for data gathering, payload distribution and deception.”
What the sources say
- Partly supportedcites Jagatic et al. 2007 sourceExperimental — accuracy not yet measured
The Figure 1 caption (section labeled Abstract; page unavailable) describes gathering public data, creating and sending a phishing email, and directing the recipient to a site that requests credentials. It numbers nine actions, however, and does not present data gathering, payload distribution, and deception as three separate steps. (LLM, conf=0.85)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Their results showed that none of these methods were effective because users would not understand or pay attention to warnings.”
What the sources say
- Not matchedcites Miller and Garfinkel 2006
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Not checked — no full text of the sourcecites Schechter et al. 2007 source
- Partly supportedcites Lin et al. 2011 sourceExperimental — accuracy not yet measured
Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Not matched
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“However, as recent research has shown [Bianchi et al. 2015], sandboxing can be effective in identifying semantic attacks that implement both behavioural and cosmetic deception.”
What the sources say
- Partly supportedcites Bianchi et al. 2015 sourceExperimental — accuracy not yet measured
Section V describes a static-analysis tool that flags apps potentially capable of GUI confusion attacks, which supports the general point that deceptive apps can be identified. The Introduction and “App classification” caution that semantic differences require human judgment. Our copy does not establish that sandboxing identifies attacks involving both behavioural and cosmetic deception. (LLM, conf=0.82)
plainstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Recently, Bianchi et al. [2015] have developed a tool for the android operating system designed to prevent malicious apps executing cosmetic (DV1), behavioural (DV2) or both cosmetic and behavioural deception (DV3).”
What the sources say
- Partly supportedcites Bianchi et al. 2015 sourceExperimental — accuracy not yet measured
The Introduction and Section V describe an Android tool that detects potentially malicious GUI-deception techniques, and Section VI describes an on-device defense that helps users identify deceptive apps. These support the claim that Bianchi et al. developed countermeasures, but not that their tool prevents malicious apps from executing the three specified categories of deception. (LLM, conf=0.86)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Crucially, in an evaluation involving 308 users, this approach was shown to significantly improve a users ability to detect malicious applications.”
What the sources say
- Partly supportedcites Bianchi et al. 2015 sourceExperimental — accuracy not yet measured
The Introduction reports a study of 308 human subjects, and the Results report significantly higher detection rates with the UI defense. The measured outcome was users’ detection of GUI confusion attacks, not their ability to detect malicious applications generally. (LLM, conf=0.85)
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
68 more claims
Create a free account to see all 198 claims and to verify the 5 references still open.
- It may state two separate claims as one. 55
- Probably not a substantial claim. 97
- The quote only partly states this claim. 1
Citations
Cites
Proceedings of the 11th USENIX Security Symposium
Boneh, D. (2002) · Cited by 2,530
Support vector machines for spam categorization
Drucker, H., Wu, D., & Vapnik, V. N. (1999) · IEEE Transactions on Neural Networks · doi:10.1109/72.788645 · Cited by 1,471
Dhamija, R., Tygar, J. D., & Hearst, M. (2006) · Proceedings of the SIGCHI Conference on Human Factors in Computing Systems · doi:10.1145/1124772.1124861 · Cited by 1,376
Kerberos: an authentication service for computer networks
Neuman, B. C., & Ts'o, T. (1994) · IEEE Communications Magazine · doi:10.1109/35.312841 · Cited by 1,337
Secrets and lies: digital security in a networked world
(2001) · Choice Reviews Online · doi:10.5860/choice.38-3941 · Cited by 1,159
Jagatic, T. N., Johnson, N. A., Jakobsson, M., & Menczer, F. (2007) · Communications of the ACM · doi:10.1145/1290958.1290968 · Cited by 1,156
Guidelines on security and privacy in public cloud computing
Jansen, W., & Grance, T. (2011) · doi:10.6028/nist.sp.800-144 · Cited by 699
(2001) · Cited by 636
Before we knew it: an empirical study of zero-day attacks in the real world
Bilge, L., & Dumitraş, T. (2012) · Proceedings of the 2012 ACM conference on Computer and communications security · doi:10.1145/2382196.2382284 · Cited by 627
Sheng, S., Holbrook, M., Kumaraguru, P., Cranor, L. F., & Downs, J. (2010) · Proceedings of the SIGCHI Conference on Human Factors in Computing Systems · doi:10.1145/1753326.1753383 · Cited by 620
Research works printed but not linked: 35
- Aburrous M., Proceedings of the 3rd International Conference on Information and Communication Technologies: From Theory to Applications (ICTTA’08)
- Algarni A., Proceedings of the 8th International Conference on Internet Technology and Secured Transactions (ICITST’13)
- Ali S. M., 2014, Integration of information security essential controls into information technology infrastructure library - A proposed framework, International Journal of Applied, 4, 1
- 10.5555/1734234.1734239 · doi:10.5555/1734234.1734239
- Bhardwaj T., Proceedings of the 3rd International Conference on Soft Computing for Problem Solving. Springer, 355--363
- Bianchi A., Proceedings of the 36th IEEE Symposium on Security and Privacy. IEEE.
- Bindra G. S., Privacy, Security, Risk and Trust (PASSAT)
- Blasing T., Proceedings of the 5th International Conference on Malicious and Unwanted Software (MALWARE). IEEE, 55--62
- 10.5555/2228340.2228358 · doi:10.5555/2228340.2228358
- Brickell E. F., 2011, Method of improving computer security through sandboxing, US Patent, 908
- Chandrasekaran M., Proceedings of the NYS Cyber Security Conference. 1--7.
- Chen T. M., 2003, Trends in viruses and worms, The Internet Protocol Journal, 6, 23
- Cluley G., 2011, Naked Security. Retrieved December 10th, 2013
- CPNI., 2013, Social Engineering: Understanding the Threat.
- Ducklin P., 2014, Even from Your Friends&excl
- Eeckhaut M., 2014, De Standaard: Belgian Professor in Cryptography Hacked.
- Emm D., 2005, Proceedings of the IWWST.
- Foozy C., Proceedings of the Malaysian Technical Universities International Conference on Engineering and Technology. 527--533
- Gavrilut D., Proceedings of the International Multiconference on Computer Science and Information Technology (IM-CSIT’09)
- Hara M., Proceedings of the IEEE Symposium on Computational Intelligence in Cyber Security (CICS’09)
- Ivaturi K., CONF-IRM Proceedings.
- Corbetta J., Proceedings of the 17th International Symposium on Research in Attacks, Intrusions and Defenses (RAID’14)
- Jordan M., Proceedings of the 14th Annual EICAR Conference. 344--364
- Konak A., Proceedings of the 2012 Networking and Electronic Commerce Research Conference.
- Lauinger T., Proceedings of the 3rd USENIX Workshop on Large-Scale Exploits and Emergent Threats (LEET’10)
- Marforio C., 2011, Report 724. Technical Report.
- Mavromatis N. P. P., 2008, USENIX Security Symposium. USENIX, 1--16
- Neupane A., Proceedings of the Network and Distributed System Security Symposium. 1--16
- Romney G. W., Proceedings of the 6th International Conference on Information Technology Based Higher Education and Training (ITHET’05)
- Rouf I., Proceedings of the 19th USENIX Security Symposium.
- Sharma V., IJCSNS(11)
- 2013, Naked Security. Retrieved
- Stewart K. E., Proceedings of the 2009 Spring Simulation Multiconference. Society for Computer Simulation International.
- D. Sullivan. 2008. What Is Search Engine Spam? The Video Edition url =. (2008). D. Sullivan. 2008. What Is Search Engine Spam? The Video Edition url =. (2008).
- Szurdi J., Proceedings of the 23rd USENIX Security Symposium (USENIX Security 14)
Cited by
Items 201–202 of 202
مهندسی اجتماعی و آسیبپذیری رفتاری کارکنان: تحلیل تهدیدات و استراتژیهای مقابله در محیطهای اداری
حیدری, س. ا. (2026) · Algorithm Scientific Research Journal · doi:10.61438/asrj.v1i1.327
Prateep, P., Gagan, V., Runwal, T., Akshitha, S., & Debnath, B. (2026) · International Conference on Innovation & Growth in Next-Gen Intelligent Technology & Engineering (IGNITE) · doi:10.1109/ignite69592.2026.11705749
“This adversary cannot attack devices or extract hardware-stored keys [27].”
Metadata
- Authors
- Heartfield, R., & Loukas, G.
- Affiliations
- University of Greenwich (GB)
- Venue
- ACM Computing Surveys
- Published
- 2015-12-09
- Volume
- 48
- Issue
- 3
- Pages
- 1-39
- ISSN
- 0360-0300 (ISSN-L)
- 1557-7341
- OpenAlex
- W2193428859
- Cited by
- 202
- References
- 153
- Citations by year
- 2026: 9 · 2025: 19 · 2024: 27 · 2023: 27 · 2022: 22 · 2021: 21 · 2020: 24 · 2019: 21 · 2018: 14 · 2017: 8 · 2016: 9
- FWCI
- 27.87
- Type
- journal article
- Language
- en
- Open access
- closed
- Locations
- ACM Computing Surveys (publishedVersion)
- Greenwich Academic Literature Archive (University of Greenwich) (acceptedVersion)
- Topics
- Physical Sciences › Computer Science › Information Systems › Spam and Phishing Detection
- Physical Sciences › Computer Science › Information Systems › Information and Cyber Security
- Physical Sciences › Computer Science › Information Systems › Web Application Security Vulnerabilities
- Keywords
- attack taxonomy, defense mechanisms, phishing emails, drive-by downloads, user deception, threat landscape, mitigation measures
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.