Sys em Secu i y Assessmen Plan Templa e
Templa e Ve sion: 1.1c DOI: 10.15497/RDA00144
Templa e Da e: No 24, 2025
P epa ed by: Pa icia Buendia , Pa ick Shi onoshi a , La s Eklund , Da id
Molik , Na alie Meye s , EOSC-Fu u e/RDA A i icial In elligence and Da a Visi a ion Wo king
G oup (AIDV WG) DV4RDA p ojec
Desc ip ion/Abs ac
This Sys em Secu i y Assessmen Plan Templa e is a s anda dized amewo k de eloped by he
ESOC Fu u e/RDA ADIV-WG and o he s o e alua e and app o e he secu i y o da a isi a ion
sys ems (i.e., pla o ms ha enable con olled, empo a y access o sensi i e esea ch da a wi hou
ans e ). The empla e equi es an ou line o eam composi ion, assessmen scope, p ocedu es, and
deli e ables o e alua ing sys em code, secu i y documen a ion, and AI model sa egua ds agains
s anda ds such as NIST SP 800-1711 and ISO/IEC 270182. The empla e also emphasizes code
e iew, au hen ica ion in eg i y, da a leakage p e en ion, dependency and con aine secu i y, and
con inuous moni o ing. These measu es suppo consis en app o al decisions, emedia ion
ecommenda ions, and inal sign-o o ensu e ha da a isi a ion echnologies mee igo ous
cybe secu i y and compliance equi emen s. This empla e is ecommended o u u e Da a Visi a ion
Sys em Secu i y Assessmen s.
Table o Con en s
Sys em Secu i y Assessmen Plan 3
1. Selec ion o an Assessmen Team by Expe ise 3
2. Scope o Assessmen and App o al C i e ia 5
Da a Visi a ion Code Re iew 5
Da a Visi a ion Sys em Secu i y Plan (SSP) Re iew 8
Da a Visi a ion Technology Re iew 9
3. Assessmen Guidelines & S anda ds 9
4. Assessmen P ocedu es 9
5. Deli e ables & Repo ing (Da a Visi a ion App o al Documen a ion) 9
6. Timeline 10
7. Use Access and Da a Flow Diag am o he Sys em o be E alua ed 11
8. Pos -Assessmen Findings and Recommenda ions 11
Appendices 12
2 ISO/IEC 27018:2025,” ISO. Accessed: No . 17, 2025. [Online]. A ailable:
h ps://www.iso.o g/s anda d/27018
1 R. Ross and V. Pilli e i, “P o ec ing Con olled Unclassified In o ma ion in Non ede al Sys ems and
O ganiza ions,” Na ional Ins i u e o S anda ds and Technology, NIST Special Publica ion (SP)
800-171 Re . 3, May 2024. doi: 10.6028/NIST.SP.800-171 3.
1
Appendix 1: Applicabili y o NIST SP 800-171 con ols o da a isi a ion sys ems 12
Appendix 2: Example Recommenda ions 14
Appendix 3: Example Flow Diag am: Use Access and Da a Flow 15
2
Sys em Secu i y Assessmen Plan
This documen ou lines he secu i y assessmen plan o a da a isi ing sys em.
Sys em Name: ____________________
Sys em Ve sion: __________
Da e Comple ed: __________
P epa ed by: ____________________
1. Selec ion o an Assessmen Team by Expe ise
To ensu e agili y and e icien ask dis ibu ion, a ocused, small eam should be o med o he
assessmen . Use he empla e o assign pe sons o essen ial oles acco ding o expe ise.
Iden i y Sys em Secu i y E alua o s o you assessmen : A eam o independen , ce i ied
secu i y p o essionals should conduc he assessmen . These expe s can be iden i ied h ough
p o essional ne wo ks, cybe secu i y i ms specializing in disciplina y speci ic domains like
heal hca e, o e e als om us ed pa ne s.
Objec i e: The p ima y objec i e is o he eam o assess and app o e he da a isi ing
pla o m's sys em secu i y implemen a ions. Ac ionable ecommenda ions may be p o ided o
u he enhance secu i y.
● Code E alua o s (Compu e Science, Cybe secu i y, and Sys em Secu i y
Backg ound): Sou ce code inspec ion should be es ic ed o indi iduals wi h a
demons able backg ound in compu e science, including ele an deg ees o subs an ial
p o essional expe ience in so wa e de elopmen and secu i y analysis. Some sys ems
equi e speci ic subspecial ies like digi al biosecu i y, o cybe biosecu i y expe s.
Task: These indi iduals will pe o m code e iews, iden i ying ulne abili ies, and
p o iding ac ionable ecommenda ions.
Requi emen s: These indi iduals mus ha e a compu e science backg ound wi h
sys em secu i y specializa ion may need o sign an NDA o access p op ie a y compu e
code.
○ Name wi h link o P o ile: _____________________________________
○ Name wi h link o P o ile: _____________________________________
○ Name wi h link o P o ile: _____________________________________
Es ima e Code E alua o s’ Time E o : (e.g. 2-3 hou s and/o 2 weeks o comple e
ask): _____________________________________
3
● Sys em Secu i y Plan (SSP) E alua o s (Sys em Secu i y Backg ound): Indi iduals
e iewing he SSP may possess a b oade ange o backg ounds, including secu i y
policy expe s, compliance o ice s, and isk managemen p o essionals.
Task: These indi iduals will e alua e he SSP o comple eness, accu acy, and
adhe ence o ele an s anda ds.
Requi emen s: These indi iduals mus be ei he secu i y policy expe s, compliance
o ice s, o isk managemen p o essionals.
○ Name wi h link o P o ile: _____________________________________
○ Name wi h link o P o ile: _____________________________________
○ Name wi h link o P o ile: _____________________________________
Es ima e SSP E alua o s’ Time E o : (e.g. 2-3 hou s and/o 2 weeks o comple e
ask): _____________________________________
● Tool E alua o s: Pilo he DV Tool (Sys em Secu i y Backg ound): Indi iduals will
ins all and un he ool.
Task: Ins all and un he da a isi a ion ool wi h sample da a.
Requi emen s: These indi iduals mus be IT p o icien .
○ Name wi h link o P ofile: _____________________________________
○ Name wi h link o P ofile: _____________________________________
○ Name wi h link o P ofile: _____________________________________
Es ima e ToolE alua o s’ Time E o : (e.g. 2-3 hou s and/o 2 weeks o comple e
ask): _____________________________________
● In e nal Liaison / Assessmen Team P ojec Manage : a designa ed ep esen a i e
om he O ganiza ion wi h a Da a Visi a ion Technology o be es ed.
Task: Ac as Assessmen Team liaison, p o iding access o necessa y documen a ion
and acili a ing p ojec communica ion be ween he assessmen eam and in e nal
s akeholde s.
Requi emen s: Backg ound in p ojec managemen o equi alen skillse
○ Name wi h link o P o ile: _____________________________________
Es ima e Liaison's Time E o : (e.g. 2-3 hou s weekly): _________________
4
2. Scope o Assessmen and App o al C i e ia
Da a Visi a ion Code Re iew
A ocused e iew o he pla o m's da a isi a ion sou ce code by quali ied indi iduals should be
unde aken o de e mine i he code mee s secu i y app o al c i e ia. The below assessmen
elemen s a e ecommended.
● Sensi i e Da a Exposu e Analysis:
○ Focus: Iden i y code segmen s ha ansmi , p ocess, o s o e sensi i e
in o ma ion (e.g., pe sonally iden i iable in o ma ion, inancial da a, au hen ica ion
c eden ials).
○ Manual Re iew: Conduc a ho ough manual e iew o code ela ed o:
■ Da a inpu and ou pu .
■ Da a s o age and e ie al.
■ API in e ac ions.
■ AJAX like calls.
○ Au oma ed Tools: (Time pe mi ing) U ilize S a ic Applica ion Secu i y Tes ing
(SAST)3 ools o au oma e code analysis o po en ial sensi i e da a leaks.
● Injec ion Vulne abili y Assessmen :
○ Focus: Iden i y and assess he isk o common injec ion a acks (e.g., SQL
injec ion, c oss-si e sc ip ing) ha could lead o da a des uc ion o unau ho ized
access.
○ Manual Re iew:
■ Examine AJAX-like calls and da abase in e ac ions o po en ial injec ion
ulne abili ies.
■ Re iew inpu alida ion and sani iza ion ou ines.
○ Au oma ed Tools:
■ Dynamic Applica ion Secu i y Tes ing (DAST)4 ools o simula e eal-wo ld
a acks.
■ In e ac i e Applica ion Secu i y Tes ing (IAST)1 ools o un ime analysis.
● Au hen ica ion and Au ho iza ion Secu i y:
○ Focus: E alua e he secu i y o use au hen ica ion and au ho iza ion
mechanisms.
○ Manual Re iew:
■ Inspec passwo d s o age and hashing p ac ices (e.g., use o sal ed
hashes).
■ Analyze API call oken gene a ion and alida ion.
■ Re iew use ole managemen and ele a ion o p i ilege ulne abili ies
(e.g., plain ex calls o igh s da abase).
○ Au oma ed Tools: (Time pe mi ing) U ilize pene a ion es ing ools o simula e
au hen ica ion and au ho iza ion a acks.
4 DAST: h ps://ieeexplo e.ieee.o g/abs ac /documen /10543484
3 SAST, IAST:
h ps://www.seman icschola .o g/pape /In e ac i e-Applica ion-Secu i y-Tes ing-Pan/4d3065d450a1
2b028b38a2e65 d5 de35537bdc8
5
● Dependency Vulne abili y Assessmen :
○ Focus: Iden i y and assess known ulne abili ies in hi d-pa y lib a ies and
packages.
○ Au oma ed Tools:
■ So wa e Composi ion Analysis (SCA)5 ools o analyze dependency
e sions and iden i y known ulne abili ies.
■ Ensu e ha all dependency lib a y/package e sions a e up o da e.
● Secu i y Moni o ing and Ale ing:
○ Focus: E alua e he e ec i eness o exis ing secu i y moni o ing and ale ing
sys ems.
○ Manual Re iew:
■ Re iew secu i y logs and moni o ing dashboa ds.
■ Assess he imeliness and accu acy o secu i y ale s.
■ Con i m he p esence o ale ing sys ems, and moni o ing o he sys ems.
●
● Consul NIST Communi y P o iles and CSWP (cybe secu i y whi e pape s)6 o
disciplina y and communi y speci ic guidance
○ Communi y P o iles p o ide a way o communi ies, (i.e., g oup o o ganiza ions
ha sha e a common con ex and an in e es in hei cybe secu i y pos u e) o
desc ibe a consensus poin o iew abou cybe secu i y isk managemen . The
NCCoE p o ides examples o Communi y P o iles and o he esou ces o help
communi ies unde s and and de elop Communi y P o iles, e.g. Cybe secu i y
F amewo k P o ile 3 o Genomic Da a (NIST IR 8467 ipd)7
○ Disciplina y-speci ic cybe secu i y whi e pape s (CSWP) o e nuanced guidance
e.g. “Cybe secu i y Th ea Modeling he Genomic Da a Sequencing Wo k low”
(CSWP) 35. This D a NIST Cybe secu i y Whi e Pape e alua es po en ial
h ea s in a genomic da a p ocessing en i onmen using an i e a i e
me hodology. I p o ides an example use case and demons a es an app oach
ha o ganiza ions can adap o iden i y cybe secu i y h ea s and mi iga ions in
hei en i onmen s.
● Res ic ed Da a Access o AI Models:
○ P inciple: Assess con ols ha p e en AI models om accessing o ans e ing
ex aneous, seconda y da a beyond he explici ly au ho ized scope
○ Ac ion:
■ Inspec da a access policies ha limi AI models o only he necessa y
da a o hei in ended unc ion.
■ De elop au oma ed es ing ools o moni o and audi da a s eams
accessed by AI models, ensu ing hey adhe e o de ined access con ols.
■ Ve i y inpu and ou pu alida ion.
■ Ve i y da a sandboxing.
● AI Model Secu i y Assessmen & Da a Leakage P e en ion:
7 h ps://doi.o g/10.6028/NIST.IR.8467.2pd
6 h ps://cs c.nis .go /publica ions/cswp
5 h ps://a xi .o g/pd /1909.00973
6
○ P inciple: Conduc ho ough secu i y assessmen s o AI models, pa icula ly
hose in e ac ing wi h ex e nal APIs (e.g., OpenAI), o iden i y and mi iga e
po en ial ulne abili ies
○ Ac ion:
■ Implemen igo ous da a leakage es ing o e i y ha sensi i e da a is no
being ansmi ed o s o ed by he AI model.
■ Assess he model's esilience agains ad e sa ial a acks, including
p omp injec ion and da a poisoning.
■ Es ablish inciden esponse plans o po en ial secu i y b eaches,
including ansomwa e a acks.
■ Implemen a e limi ing o p e en abuse and educe he isk o da a
ex il a ion h ough epea ed o au oma ed que ies.
● Secu ing In elligen Agen s
○ As AI sys ems e ol e om simple assis an s o ully au onomous agen s, hey
in oduce inc easingly complex secu i y isks ha da a isi a ion pla o ms using
such agen s mus ac i ely add ess. Each le el o au onomy, om obse ing and
ac ing unde human guidance o making independen decisions, c ea es new
ulne abili ies ha adi ional inciden esponse s a egies a e no equipped o
manage. Gi en he signi ican compu a ional demands o AI agen s, da a
isi a ion pla o ms a e gene ally expec ed o ely on no mo e han a single agen .
Howe e , e en one agen , when g an ed decision-making au ho i y and ool
access, p esen s a delica e balance be ween u ili y and isk.
○ The Coali ion o Secu e AI (CoSAI)’s AI Inciden Response F amewo k8 is
add essing hese unique, dynamic isks posed by in elligen , au onomous
sys ems wi h suppo om majo echnology companies, including Google, which
ecen ly dona ed da a om i s secu e AI amewo k (SAIF)9.
○ Key assessmen p ocedu es ecommended by CoSAI include:
■ Agen Beha io P o iling: Es ablish baselines o no mal agen beha io o
de ec anomalies
■ Tool In oca ion Audi ing: Log and e iew all ool calls made by agen s o
misuse o escala ion
■ P omp and Ou pu Sc eening: Analyze inpu s and ou pu s o signs o
injec ion, hallucina ion, o leakage
■ C oss-Team Inciden Playbooks: De elop sha ed p o ocols ac oss
secu i y, ML, and p oduc eams o coo dina ed esponse
■ Red Teaming and Simula ion: Regula ly es agen esilience h ough
ad e sa ial scena ios and s ess es ing
■ Pos -Inciden Fo ensics: Conduc oo cause analysis and upda e
gua d ails and policies based on indings
9
h ps://www.oasis-open.o g/2025/09/16/google-dona es-secu e-ai- amewo k-sai -da a- o-coali ion-
o -secu e-ai/
8
h ps://gi hub.com/cosai-oasis/ws2-de ende s/blob/main/inciden - esponse/AI%20Inciden %20Resp
onse.md
7
■ Con inuous Risk Mapping: Use dynamic h ea models ha e ol e wi h
agen capabili ies and deploymen con ex s
● Da a Co up ion P e en ion & Model Ve sioning:
○ P inciple: E alua e measu es used o p e en da a co up ion and ensu e da a
in eg i y, pa icula ly o ime-sensi i e o highly sensi i e da a (e.g., genomic
da a).
○ Ac ion:
■ Ve i y ha obus da a in eg i y checks ha e been implemen ed, including
checksums and da a alida ion ou ines, h ough code inspec ion and
e iew.
■ U ilize "Sa e Ou S a e" (SOS) es s o iden i y and p e en da a
co up ion.
■ Ve i y ha s ic model e sioning and locking mechanisms ha e been
implemen ed, o p e en unau ho ized modi ica ions o da a d i .
■ Ve i y ha da a p o enance acking has been implemen ed, and e i y
ha acking is accu a e
● Tempo ali y and Time D i Managemen :
○ P inciple: Add ess he challenges o da a empo ali y and ime d i , which can
impac he accu acy and eliabili y o AI models.
○ Ac ion:
■ Ve i y ha imes amping and e sioning mechanisms o all da a ha e
been implemen ed.
■ Assess da a d i de ec ion and mi iga ion s a egies.
■ Assess ime-sensi i e da a alida ion echniques.
Da a Visi a ion Sys em Secu i y Plan (SSP) Re iew
This Assessmen plan empla e depends on a en ion o he ele an NIH equi ed NIST SP
800-171 secu i y con ols implemen ed in he da a isi ing sys em(s) being es ed.
Rele an sec ions o he SSP should be e alua ed o app o al based on hei comple eness,
accu acy, and adhe ence o s anda ds speci ic o da a isi a ion secu i y. The SSP should
add ess he p o ec ion o Con olled Unclassi ied In o ma ion (CUI) as de ined by NIST SP
800-171, whe e applicable, conside ing he limi ed scope o da a ansmission.
Applicable Con ols: Fo da a isi a ion scena ios, NIST SP 800-171 ules conce ning access
con ol, au hen ica ion, and au ho iza ion a e highly applicable as hey go e n who can access
he da a isi a ion ool and he unde lying da a. Audi and accoun abili y ules a e also ele an
o acking ac ions pe o med du ing da a isi a ion sessions. Rules ega ding sys em and
in o ma ion in eg i y, including p o ec ion agains malwa e, a e c ucial o ensu e he da a isi ing
ool doesn' comp omise he hos sys em o he da a i sel .
No applicable Con ols: Con e sely, ules a ound physical p o ec ion, media p o ec ion
(pe aining o physical media), and a guably ha dwa e main enance ha e limi ed di ec
applicabili y o he da a isi a ion p ocess i sel , especially when he da a isi a ion ool is a
empo a y so wa e laye ins alled on ha dwa e no owned o main ained by he da a isi o ; he
secu i y o he unde lying ha dwa e emains he esponsibili y o he da a owne . Howe e , he
8
da a owne 's adhe ence o all NIST SP 800-171 con ols is s ill pa amoun o he o e all
secu i y o he CUI being isi ed.
Fo u he in o ma ion, consul Appendix 1: Applicabili y o NIST SP 800-171 con ols o da a
isi a ion sys ems.
Da a Visi a ion Technology Re iew
The co e echnologies enabling da a isi a ion can be es ed wi h sample da a o assess
secu i y obus ness and compliance wi h ele an s anda ds using he below Assessmen
Guidelines and S anda ds, and ollowing he ecommended Assessmen P ocedu es.
3. Assessmen Guidelines & S anda ds
● NIH Guidance o NIH Con olled-Access Da a: NIH uses he NIST SP 800-171
guidance. The assessmen should e i y compliance wi h he con ols ou lined in NIST
SP 800-171, pa icula ly hose ela ed o da a access and secu i y, as a key equi emen
o da a isi a ion ea u e app o al.
● Da a Visi a ion Indus y Bes P ac ices: The assessmen should conside indus y
bes p ac ices o secu e da a access and isi a ion as benchma ks o app o al.
● ISO/IEC s anda d ega ding Cloud and PII: In o ma ion echnology – Secu i y
echniques – Code o p ac ice o p o ec ion o pe sonally iden i iable in o ma ion (PII) in
public clouds ac ing as PII p ocesso s (ISO/IEC 27018:2019)
● The Coali ion o Secu e AI (CoSAI) p oposes a new, AI-speci ic inciden esponse
amewo k ha emphasizes con inuous moni o ing, c oss- unc ional collabo a ion, and
apid con ainmen s a egies ailo ed o he luid na u e o AI agen h ea s.
4. Assessmen P ocedu es
● Da a Visi a ion Code Re iew: Quali ied compu e scien is s will conduc ocused code
analysis o he da a isi a ion ea u e o in o m he app o al decision.
● Da a Visi a ion Policy and P ocedu e Re iew : The ele an sec ions o he SSP and
ela ed documen a ion will be e iewed o da a isi a ion app o al compliance.
● Da a Visi a ion Da a Flow Analysis : Da a low analysis speci ic o he da a isi a ion
ea u e will in o m he app o al decision.
● Da a Visi a ion Secu i y Con igu a ion Re iew: Secu i y con igu a ion e iews o he
da a isi a ion in as uc u e will in o m he app o al decision.
5. Deli e ables & Repo ing (Da a Visi a ion App o al
Documen a ion)
Fo each lis i em below assign a esponsible pa y and eco d when he epo /ac i i y was
comple ed and how o access i :
9
b. No e: Login is equi ed o he QC-App IP add ess o localhos sepa a ely
om FAIRLYZ.com
c. No e: login is no equi ed i al eady logged in. The session expi es a e 24
hou s.
S ep 6. The sys em e i ies ha he use who logs in is he same use who owns he s udy
and da a p o ile in FAIRLYZ.com
a. I e i ied, he sys em p oceeds wi h da a isi a ion and QC
b. I no e i ied, he access is blocked
S ep 7. Da a access is es ic ed o he Docke ins alla ion olde .
S ep 8. The use uns QC on he da a
S ep 9. The use syncs QC esul s o FAIRLYZ.com
16
Please ci e his documen as: Pa icia Buendia, Pa ick Shi onoshi a, La s Eklund, Da id
Molik and Na alie Meye s. “Sys em Secu i y Assessmen Plan Templa e.” DV4RDA P ojec o
he EOSC-Fu u e/RDA A i icial In elligence and Da a Visi a ion Wo king G oup. Resea ch
Da a Alliance. No embe 17, 2025.
17
Acknowledgemen s
This DV4RDA p ojec has ecei ed unding h ough RDA TIGER om he
Eu opean Union’s Ho izon Eu ope amewo k p og amme unde g an
ag eemen No. 101094406. Views and opinions exp essed a e howe e
hose o he au ho s only and do no necessa ily e lec hose o he
Eu opean Union o ins i u ions ep esen ed he e. Nei he he Eu opean
Union no he ins i u ions can be held esponsible o hem.
