Page 194 - Cyber Defense eMagazine July 2024
P. 194

1.  The  LlamaV2  7B  model  is  highly  susceptible  to  both  direct  and  indirect  Prompt  Injection  (PI)
                   attacks, with a majority of test attacks succeeding when exposing the model to contexts containing
                   injected prompts.
               2.  The model is vulnerable to Adversarial Jailbreak attacks, provoking responses that violate ethical
                   guidelines,  with  tests  revealing  a  significant  reduction  in  the  model's  refusal  rate  under  such
                   scenarios.
               3.  The  model  is  highly  susceptible  to  Denial-of-Service  (DoS)  attacks,  with  prompts  containing
                   transformations  like  word  replacement,  character  substitution,  and  order  switching  leading  to
                   excessive token generation.
               4.  The model demonstrateד  a high propensity  for data leakage  across diverse  datasets, including
                   finance, health, and generic PII.
               5.  The model has a significant tendency to hallucinate, challenging its reliability.
               6.  The model often opts out of answering questions related to sensitive topics like gender and age,
                   suggesting  it  was  trained  to  avoid  potentially  sensitive  conversations  rather  than  engage  with
                   them in an unbiased manner.




            DeepKeep’s  evaluation  of  data  leakage  and  PII  management  demonstrates  the  model's  struggle  to
            balance user privacy with the utility of information provided. However, Meta’s LlamaV2 7B LLM shows a
            remarkable ability to identify and decline harmful content, boasting a 99% refusal rate in our tests. Yet,
            our investigations  into hallucinations  indicate a significant tendency to fabricate responses,  challenging
            its  reliability.  Overall,  the  LlamaV2  7B  model  showcases  strengths  in  task  performance  and  ethical
            commitment,  with  areas  for  improvement  in  handling  complex  transformations,  addressing  bias,  and
            enhancing security against sophisticated threats.


            Dr. Rony  Ohayon  is  the  CEO  and  Founder  of Deep-
            Keep,  the  leading  provider  of  AI-Native  Trust,  Risk,
            and Security Management  (TRiSM). He has 20 years
            of experience  within the high-tech industry with a rich
            and  diverse  career  spanning  development,  technol-
            ogy, academia, business, and management. He has a
            Ph.D.  in  Communication  Systems  Engineering  from
            Ben-Gurion  University,  a  Post-Doctorate  from  ENST
            France, an MBA, and more than 30 registered patents
            in  his  name.  Rony  was  the  CEO  and  Founder  of
            DriveU,  where  he  oversaw  the  inception,  establish-
            ment,  and management.  Additionally,  he  founded  Li-
            veU,  a  leading  technology  solutions  company  for  broadcasting,  managing,  and  distributing  IP-based
            video content,  where he also served as CTO until the company  was acquired.  In the education  realm,
            Rony was a senior faculty member at the Faculty of Engineering  at Bar-Ilan University (BIU), where he
            founded the field of Computer Communication  and taught courses about algorithms, distributed compu-
            ting, and cybersecurity in networks.








            Cyber Defense eMagazine – July 2024 Edition                                                                                                                                                                                                          194
            Copyright © 2024, Cyber Defense Magazine. All rights reserved worldwide.
   189   190   191   192   193   194   195   196   197   198   199